You are on page 1of 353

MELANCHOLIA: A DISORDER OF MOVEMENT AND MOOD

It has long been accepted that depressive disorders comprise a


biologically based type - the so-called 'endogenous' or 'melancholic'
depression - and a residual set of depressive conditions resulting from
social factors. The difficulty has been in distinguishing the melancholic
type of depression on the basis of clinical features.

In the belief that melancholia is capable of clinical definition, and that


psychotic depression may also be clinically defined, the authors
describe in this book the results of research studies seeking to
distinguish those conditions from residual and more heterogeneous
depressive disorders. They describe the development of a behavioral
sign-based approach, the CORE system, and demonstrate its
superiority to previous symptom-based diagnostic systems. The CORE
measure also identifies neurobiologically discrete groups among
depressive disorders, and the authors suggest that the psychomotor
signs elicited may indicate the likely pathogenesis of melancholic
depression, involving the basal ganglia and connections to the frontal
cortex.

This is therefore a challenging new account of the classification and


neurobiology of depression, one that is certain to interest all clinicians
involved in the evaluation or treatment of such patients. In addition to
reviewing the historical and neurobiological evidence for melancholia as
a neuropsychiatric disorder, it provides all the necessary details of the
CORE diagnostic system, and the CORE measure itself is incorporated
as an appendix.
MELANCHOLIA:
A DISORDER OF MOVEMENT
AND MOOD
A Phenomenological and Neurobiological Review

Edited by
GORDON PARKER and DUSAN HADZI-PAVLOVIC
School of Psychiatry, University of New South Wales, Sydney, Australia

with the editorial assistance of


KERRIE EYERS
Mood Disorders Unit, Prince Henry Hospital, Sydney, Australia

CAMBRIDGE
UNIVERSITY PRESS
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 1RP
40 West 20th Street, New York, NY 10011-4211, USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia

© Cambridge University Press 1996


First published 1996

A catalogue record of this book is available from the British Library.

Library of Congress cataloguing in publication data applied for.

ISBN 0-521-47275-X

Transferred to digital printing 2004


Contents

List of contributors page vii


Acknowledgments ix
Introduction G. Parker and D. Hadzi-Pavlovic 1

Part One: Classification and Research:


Historical and Theoretical Aspects
1 Issues in Classification: I. Some Historical Aspects P. Boyce
and D. Hadzi-Pavlovic 9
2 Issues in Classification: II. Classifying Melancholia G. Parker,
D. Hadzi-Pavlovic and P. Boyce 20
3 Issues in Classification: III. Utilising Behavioural Constructs
in Melancholia Research I. Hickie 38
4 Issues in Classification: IV. Some Statistical Aspects
D. Hadzi-Pavlovic 57

Part Two: Development and Validation of a Measure of


Psychomotor Retardation as a Marker of Melancholia
5 Psychomotor Change as a Feature of Depressive Disorders:
Historical Overview and Current Assessment Strategies
G. Parker and H. Brotchie 67
6 Development and Structure of the CORE System G. Parker
and D. Hadzi-Pavlovic 82
7 Reliability of the CORE Measure D. Hadzi-Pavlovic and
G. Parker 130
8 Validity of the CORE: I. A Neuroendocrinological Strategy
P. Mitchell 138
9 Validity of the CORE: II. Neuropsychological Tests /. Hickie 149
vi Contents
10 Validity of the CORE: III. Outcome and Treatment Predic-
tion G. Parker, I. Hickie and C. Mason 160
11 Phenotypic Expression of Melancholia Contrasted for Those
with Bipolar and Unipolar Illness Courses P. Mitchell and
A. Sengoz 172
12 Psychotic Depression: Clinical Definition, Status and the
Relevance of Psychomotor Disturbance to Its Definition
G. Parker, I. Hickie and D. Hadzi-Pavlovic 179
13 A Clinical Algorithm for Defining Melancholia: Comparison
with Other Sub-typing Measures G. Parker and D. Hadzi-
Pavlovic 202
14 Rating the CORE: A User's Guide K. Wilhelm 211

Part Three: The Neurobiology of Melancholia


15 Melancholia as a Neurological Disorder M.-P. Austin and
P. Mitchell 223
16 Melancholia and the Ageing Brain H. Brodaty 237
17 Magnetic Resonance Imaging in Primary and Secondary
Depression I. Hickie, C. Hickie, E. Scott and K. Wilhelm 252
18 Functional Neuroimaging in Affective Disorders M.-P. Austin
and P. Mitchell 267
19 Summary and Conclusions G. Parker and D. Hadzi-Pavlovic 277

The CORE Measure: Procedural Recommendations and


Rating Guidelines 282
References 292
Author Index 324
Subject Index 335
Contributors

Marie-Paule Austinf MB BS FRANZCP


Staff Specialist in Psychiatry, Prince Henry Hospital. MD U Consultant
Philip BoyceJ LRCP MRCS MB BS MD FRANZCP DIP. PSYCHOTHERAPY
Professor, University of Sydney and Department of Psychiatry, Nepean
Hospital. MDU Consultant.
Henry Brodatyf MB BS FRACP FRANZCP MD
Professor, Department of Psychogeriatrics, The Prince Henry Hospital
and University of New South Wales, Prince Henry Hospital
MDU Consultant.
Heather Brotchie| MB BS
MDU Research Assistant.
Kerrie Eyersf MA DIPED. MPH
MDU Administrative Co-ordinator.
Dusan Hadzi-Pavlovicf BSC(PSYCHOL) MPSYCHOL
Senior Hospital Scientist, Division of Psychiatry and Adjunct Senior
Lecturer, School of Psychiatry, University of New South Wales, Prince
Henry Hospital. MDU Co-investigator.
Catherine HickieJ MB BS FRANZCP
MDU Associate Investigator.
Ian Hickief MB BS MD FRANZCP
Associate Professor, School of Psychiatry, University of New South
Wales, St. George Hospital. MDU Consultant.
Catherine Masonf MB BS MPH FRANZCP
Senior Lecturer, School of Psychiatry, University of New South Wales,
Prince of Wales Hospital. MDU Associate Investigator.
f Current affiliate of the Mood Disorders Unit
X Previous affiliate of the Mood Disorders Unit

Vll
viii Contributors

Philip Mitchellf MB BS MD F R C P S Y C H FRANZCP


Associate Professor, School of Psychiatry, University of New South
Wales, Prince Henry Hospital. MDU Consultant
Gordon Parkerf MB BS MD P H D FRANZCP
Professor, School of Psychiatry, University of New South Wales, Prince
of Wales Hospital. MDU Consultant.
Elizabeth Scottf MB BS BSc
Research Fellow, New South Wales Institute of Psychiatry. MDU Asso-
ciate Investigator.
Ayse SengozJ BSc
MDU Research Assistant.
Kay Wilhelmf MB BS MD FRANZCP
Staff Specialist in Psychiatry, Prince Henry Hospital. MD U Consultant.
Acknowledgments

The research reported in this book has been largely funded by the Na-
tional Health and Medical Research Council of Australia. Administra-
tive and financial backing were also provided by the Prince Henry and
Prince of Wales hospitals. We are extremely grateful for that support.
Our thanks to Sandra Evans, Yvonne Foy, Zora Vuckovic and Lynette
Campbell for their careful and invaluable preparation of this manuscript
and of the many Mood Disorders Unit (MDU) papers contributed to this
volume.
To Chris Taylor for her dedicated quality control over data collection
and entry.
To the clinical staff who have managed the MDU patients and the
many practitioners who have both referred patients to the MDU and
assisted with data collection.
To the patients who have contributed so willingly and helpfully to
many studies.
To the editors and publishers of the Australian and New Zealand Jour-
nal of Psychiatry, Biological Psychiatry, British Journal of Psychiatry,
Journal of Affective Disorders, Journal of Nervous and Mental Disease
and Psychological Medicine for permission to abstract from journal ar-
ticles.
The manuscript was prepared using I&TjgX, and we thank the often
invisible community members who provide the tools that make up this
essentially public domain system.

IX
Introduction
GORDON PARKER
DUSAN HADZI-PAVLOVIC

It may then be useful to investigate two or three symptoms of depression as


a means of discovering the necessary biological processes responsible for their
emergence and contribution to the syndrome of depression. This is likely to be
the wish of the psychopathologist who seeks to understand the fundamental
nature of depression. - Costello (1993)

More than a quarter of a century ago, Carney, Roth and Garside


(1965) wrote that the "establishment of a classification of affective dis-
orders commanding wide agreement among clinical practitioners and
investigators is one of the most pressing needs of contemporary psychia-
try." We suggest that that need persists, and that the continuing lack of
such a classification has both restricted and slowed pursuit of aetiologies
and refinement of treatment approaches. Kendell (1989) has provided
some historical analogies:
It was only after Sydenham had demonstrated that "the pox" was actually
two distinct syndromes, chicken pox and small pox, that it was possible to
predict with any accuracy who would almost certainly recover and who would
remain scarred for life and was in danger of dying. And only after physicians
had learned to distinguish between the renal and cardiac forms of dropsy was
it possible to predict which patients were likely to benefit from digitalis.

An ongoing objective - that has alternately excited and frustrated


researchers - has been to distinguish an assumed categorical depressive
entity, with an imputed preferential neurobiological base. Any such en-
tity has variably been termed endogenous, psychotic, vital, autonomous,
endogenomorphic, Type A and melancholic depression (see Andreasen
1980; Klein 1974), and we now adopt the last term. To the binarians,
melancholia is a categorical entity; to the Unitarians, depressive disorders
vary dimensionally, with melancholia viewed merely as a more severe ex-
pression. An extended debate between advocates of each view occurred

1
2 G. Parker and D. Hadzi-Pavlovic
in the sixties and seventies. Limitations to the multivariate analyses then
available led to inconclusive results. Failure to provide firm proof of the
binary hypothesis both discouraged interest in attempts at further reso-
lution and led to dimensionally weighted measures of melancholia being
respected in subsequent DSM and ICD classificatory systems. We seek to
revive interest in the clinical identification of melancholia (whatever its
status), and we declare our binarian bias from the outset.
Blumenthal (1971) held that depressed patients may manifest hetero-
geneity across three domains (i.e., aetiology, clinical features and out-
come - with the last including response to treatment). Ni Brolchain
(1979) argued, however, for classification being based in the first in-
stance on mental state items. Once this classification is achieved, the
researcher may proceed to examine how aetiological and background fac-
tors are associated with clinically defined sub-types, so that "more spe-
cific questions can be investigated and more precise answers obtained."
In the same way that dementia is not viewed as a single entity, but
as a broad class comprising heterogeneous conditions with varying aeti-
ologies (e.g., Alzheimer's type, vascular type, alcohol-induced), clinical
features and outcome, we believe that there is a need to determine clin-
ically meaningful depressive sub-types. Believing that melancholia is
capable of clinical definition, and that psychotic depression may also be
clinically defined (whether as a sub-type of melancholic depression or
as a separate sub-type), we have undertaken a series of research studies
seeking to distinguish clinically those conditions from residual and more
heterogeneous depressive disorders.
We detail our attempts at clinical definition (and several validation
studies) in this book. Our key findings are readily summarised. First,
clinical definition appeared superior using a sign-based approach, so
that our final measure (the CORE) differs from previous work and cur-
rent international diagnostic instruments which have a symptom focus.
Secondly, the CORE measure appears to identify domains of psychomo-
tor disturbance that seem quite specific to psychotic and non-psychotic
melancholia - so allowing allocation of depressed patients to putative
melancholic and non-melancholic depression classes. Thirdly, as clinical
features are simply the surface (or recordable) markers of underlying
neuropathological processes, the CORE measure potentially allows iden-
tification of neurobiologically discrete groups. Not only does this assist
aetiological and treatment research, it points at the likely sites of the
neurobiological perturbations and/or lesions contributing to melancholic
depression. The identified specificity or CORE-measured psychomotor
Introduction 3
disturbance to melancholia was not anticipated, as behavioural expres-
sions of psychiatric disorders are rarely typical of or specific for psychi-
atric conditions, with van Praag (1993) noting that much of the "biol-
ogy" uncovered in mental disorders appears "to be devoid of diagnostic
specificity." Additionally, he argued that as most disease entities are, on
the whole, "pseudoentities, it would be little less than a miracle if ... a
single marker ... would be found identifying such an 'entity.'"
These several points have been considered more generally by Carpen-
ter et al. (1993) in relation to schizophrenia. Those authors advocated
identifying physiological correlates and, by so reducing syndromal het-
erogeneity, defining a sub-group homogeneous for the marker of inter-
est. They argued for "starting with the observable clinical presentation
of the patient," grouping clinical features into theoretically important
categories and then searching for the underlying pathological processes.
They suggested that the intent should then be "to shift the focus of
investigations to clinicopathologic correlations of specific psychopatho-
logic domains with discrete neural circuits." Thus, they emphasised the
need for researchers to move beyond a hypothesis-generating mode to
a hypothesis-testing framework. To do that, classificatory approaches
must provide an adequate foundation for what the authors described as
"strong inference research." If current classificatory systems (i.e., those
providing criteria for a diagnostic category known as melancholia) fail
to provide such a foundation, as we argue, then the focus should turn
to markers of the neuropathological processes underlying melancholia.
Our book has therefore been framed by both the results and the im-
plications of our research, with the latter consonant with the Carpen-
ter model. In arguing that melancholia is imprecisely defined by many
melancholia measures and indices, we review (Part I) relevant theoret-
ical, historical and statistical approaches to its classification. As our
empirical research has the potential to revive interest in rating clinical
signs of psychomotor disturbance, we also provide an historical overview
of the role of psychomotor change in depression.
In Part II we describe the progressive development of the final CORE
measure of clinician-rated psychomotor disturbance, consider the extent
to which the measure may identify a domain which is "necessary and suf-
ficient" to valid definition of melancholia, develop an empirical algorithm
and contrast its properties with several current measures of melancho-
lia, and report a number of studies considering properties of the CORE
measure, especially its reliability and validity. The latter studies suggest
that deriving homogeneous samples of depressed subjects on the basis
4 G. Parker and D. Hadzi-Pavlovic
of CORE scores appears to generate a considerable increase in the speci-
ficity of neurobiological findings, a general proposition best illustrated
by reference to dexamethasone suppression and neurocognitive testing.
Although we are unclear as to the status of psychotic (or delusional) de-
pression - whether it is a distinct condition or a melancholic sub-type -
we also report attempts at its clinical definition, and we demonstrate the
striking relevance of CORE-measured psychomotor disturbance to that
disorder as well.
In Part III we explore a number of implications of the demonstrated
specificity of psychomotor disturbance to melancholia. At the simplest
level, we suggest that as melancholia is as much a disturbance of move-
ment as of mood, then its status as a psychiatric disorder versus a neuro-
logical disorder (or at least a neuropsychiatric disorder) is worthy of con-
sideration. We suggest that the specificity of psychomotor disturbance
provides information on likely underlying neurobiological processes as
well as the likely psychopathological domains of interest (i.e., striatal,
limbic and prefrontal regions and their interconnections). We recognise
that determinants may be either structural and/or functional and review
relevant imaging studies, as well as the effects of "the ageing brain" and
other influences of possible relevance to any pursuit examining the neu-
robiological basis to melancholia.
While van Praag (1993) argued that psychopathological (or clinical)
features are rarely specific for a syndrome or nosological entity, perhaps
explaining "the failure to find specific markers of psychiatric disorders,"
he did suggest that "functionally oriented psychopathology" would bene-
fit from measurement of abnormal behaviour. His pessimism was clearly
driven by the reality that there are few homogeneous psychiatric entities.
We believe, however, that melancholia is one such entity and capable of
being clinically circumscribed. If this is true, the possibility of identi-
fying melancholia by a biological marker is ever present. Others clearly
have had the same belief, with markers such as the dexamethasone sup-
pression test, low platelet monoamine oxidase, phase advance variables
as well as other tests being proposed. We offer the CORE system then
as another biological marker - a measure that is weighted to clinical do-
mains that appear specific to melancholic depression and that therefore
allows melancholic and non-melancholic depression to be distinguished.
When we established the Mood Disorders Unit (MDU), we set as our
initial research task the clinical distinction of melancholic depression -
although our motivation was possibly driven as much by what Hinshel-
wood described as "the adventure of discovery itself." All contributors
Introduction 5
to this book have worked on that task at the MDU, and we hope that
our shared involvement and enthusiasm in that initial objective provide
a cohesiveness to this book.
Part One
Classification and Research:
Historical and Theoretical Aspects
1
Issues in Classification:
I. Some Historical Aspects
PHILIP BOYCE
DUSAN HADZI-PAVLOVIC

Introduction
The classification of the various types of depression - if types there be -
is still to be resolved. Some proposed classifications have been widely ac-
cepted: the unipolar-bipolar dichotomy, and the removal of involutional
melancholia from official classification systems in particular (Farmer and
McGuffin 1989). Others, however, have not. In the case of melancholia,
for example, the last 15 years have seen major changes in the consensus
definition developed for the official nomenclature of DSM-m, DSM-III-R,
and DSM-IV. Indeed the concept of melancholia as an entity is not univer-
sally accepted, as demonstrated by ICD-10 (World Health Organization
1992) preferring the term "somatic" and acknowledging that "the scien-
tific status of this syndrome is in any case questionable."
The issue which has remained contentious, which in many ways drives
the classification debate, and which is important to the theme of this
book, is the relationship between the various forms of depression: in
particular, the relationship between a genetically determined "constitu-
tional" depression and those depressions (or that depression) which arise
from psychosocial determinants. In British psychiatry the ground rules
for the debate were set during the 1920s and 1930s by a handful of influ-
ential clinical studies which argued either for some form of dichotomy or
for a unitary view of depression in which the various forms of depression
lay along a continuum.
10 P. Boyce and D. Hadzi-Pavlovic
Background
In the 1920s, when the debate over depression classification in British
psychiatry started (or restarted), there were a number of shaping influ-
ences.
(i) The majority of persons with depression were treated in mental in-
stitutions. Diagnostic decisions about depression centered on the
issue of certification - patients suffering from manic depressive ill-
ness were certified and placed in an institution. In such institutions
Kraepelin's formulations (see below) appeared salient, as his work
was based upon patients seen in mental institutions, and not on
the less severe, ambulatory, depressed patients who usually did not
come to the attention of psychiatrists.
(ii) In psychiatric nomenclature, the end of the 19th century had seen
the 5th (1896) and 6th (1899) editions of Kraepelin's textbook,
which made the important distinction between dementia praecox
and manic depressive insanity, with the latter described as "manic
excitement ... or ... a peculiar depression with psychomotor inhi-
bition, or a mixture of the two states." Kraepelin considered manic
depressive illness to arise as a result of a defective constitution and
so gave it a clear biological basis.
(iii) The recognition and mangement of neurotic disorders changed with
the growth of the psychoanalytic movement (and psychotherapy
more generally following the lessons learnt dealing with the psycho-
logical sequelae of shell shock during World War I [Merskey 1991]).
Psychoanalysis extended to influence the treatment of manic de-
pressive illness and, importantly, added psychodynamic issues to
the aetiological equation (e.g., Abraham 1911).
(iv) Psychiatry absorbed the aetiological implications of the post-ence-
phalitc neuropsychiatric syndromes. Interestingly, it was the symp-
toms observed in enchephalitis lethargica that prompted Naville to
coin the word bradyphrenia (Rogers 1992). As discussed in later
chapters such mental slowing is one feature that unites Parkinson's
disease (PD) and melancholia.
(v) Psychotherapy became accepted as a treatment method for neur-
asthenia, a disorder which embraced what would now be called (us-
ing DSM-III et seq) major depression, generalised anxiety disorder
and somatisation.
(vi) Psychiatric practice extended beyond the confines of the asylum
(where manic depressive insanity was over-represented in the class
Some Historical Aspects 11
of depression) to office-based psychiatry (where neurasthenia was
over-represented).
Clinically, if not taxonomically, a resulting dichotomy in the approach
to depressed patients was summarised by the axiom "analyse a neurotic,
guard a psychosis; don't guard a neurosis, don't analyse a psychosis; I
cure the neurotic, time cures the psychotic" (Curran 1937).

The 1926 Meeting


In 1926 Edward Mapother, the then medical superintendent of the Maud-
sley Hospital, delivered a paper to the British Medical Association on
manic depressive illness (Mapother 1926). This paper challenged the
prevailing views of Kraepelin on the distinction between psychoses and
neuroses with reference to depression, which, he noted, "has really grown
out of practical differences, particularly as regards certification and asy-
lum treatment." Unable to determine any differences between these
conditions with regard to causation, prognosis or treatment Mapother
considered it pointless to distinguish between them. He went on to point
out that both neuroses and psychoses can be precipitated by stress (a
challenge to the endogenous nature of psychoses). Most importantly,
he observed that there is "a complete graduation (from constitutional
symptoms) to certain minor disturbances which, whether continuous or
physic, are usually termed neuroses."
Mapother's paper was the starting point for the unitary view of de-
pression in the United Kingdom. (In the United States the Meyerian
view of depression largely prevailed. It was dominated by the notion
of reaction types and included a dimensional model of psychological ill-
ness.) Mapother's views were immediately challenged in the discussion
that followed. First, Buzzard asserted that "the distinction between
these two disorders, although sometimes difficult, was of fundamental
importance in regard to prognosis and treatment, as well as in academic
accuracy." Ross followed up by stating, "If Dr. Mapother would carefully
study mental states he would find they would lead him to perceive fun-
damental differences between the psychoses and psychoneuroses." Ross
went on to point out "some important differences" between the depres-
sive psychotic and the depressive psychoneurotic. One in particular was
the reaction to the environment, with the psychotic held to be unable
to gain pleasure from any positive event (unreactive); a view supported
by another discussant, Gillespie.
12 P. Boyce and D. Hadzi-Pavlovic

The Binary View of Depression


The champion of the binary view was Gillespie, co-author with Hen-
derson of one of the most influential textbooks in British psychiatry
(Henderson and Gillespie 1927). In this textbook a distinction between
the depressive types, based on Henderson's research, was clearly made.
In 1929, Gillespie published his study of 25 patients admitted to the
Cassel Hospital. In this paper he abandoned the endogenous-reactive
dichotomy by shifting away from aetiology as the basis of classification
and used instead the course of the patient's illness as the basis for clas-
sification. He proposed two types of depression. First, he defined an
autonomous type, which, once precipitated, ran an independent course
and was unresponsive to the environment. This type had essentially
been previously described as endogenous depression under Kraepelin's
classification. Secondly, he proposed a reactive type, in which the depres-
sion was responsive to the environment. This use of the term "reactive"
was strikingly different from the way it had been applied earlier. Gille-
spie did not use it to indicate psychogenesis, as others had done, but
rather to describe the patient's response to the environment. He viewed
the autonomous group as heterogeneous and identified two subgroups:
a pure group, and a smaller group with a clinical picture consistent with
descriptions of involutional melancholia. For Gillespie, the attraction of
these two types was parsimony; he pointed out that there were a number
of other ways of grouping the same patients.
It is worth noting that Gillespie, Henderson and Lewis were all stu-
dents of Adolph Meyer. Lewis described an apostate Gillespie who was
"professedly a follower of Adolph Meyer, [but who] has manifestly been
greatly influenced by the nosological views of Kraepelinians ..." (Lewis
1934).

The Unitary View


The proponents of the unitary view of depression carried out clinical
studies as thorough as Gillespie's, but came to very different conclusions.

Curran. Curran (1937) assessed a group of 25 depressives referred to


the Hospital for Epilepsy and Paralysis in Maida Vale, England. He
used 11 clinical features - modified from Gillespie's work - which were
considered to distinguish the pattern of illness associated with psychotic
or neurotic depressive types. After rating each patient on these criteria,
Some Historical Aspects 13
Curran was unable to identify distinct groups of patients corresponding
to the two types. He came to the very Meyer-like conclusion that there
was no sharp distinction between the neuroses and manic depressive psy-
chosis: "It is argued that it [the distinction] does not in fact exist, and
that its attempted erection results from a lingering formalism based upon
the conception of 'disease entities'; whereas the broader conception of
'reaction types' fits the facts to be observed in a more satisfactory way."
Later, Curran and Mallinson (1941) assessed 88 consecutive depressed
cases admitted to a Royal Navy auxilliary hospital. The patients were
categorised into endogenous and reactive types and rated on nine his-
torical and 11 symptom variables. Again, the depressive types could not
be distinguished by either historical variables or symptom pattern. Im-
portantly, Curran and Mallinson could not identify any differences with
regard to aetiology, especially constitutional (family history of mental
illness); physical factors; environmental factors (life stress - disruption
to home life, domestic disturbance or exposure to danger); and per-
sonality factors. While they conceded that some cases met the criteria
for endogenous or neurotic depression, they noted that "such cases, how-
ever, are not separated by a sharp borderline but by a wide borderland."
Finding no clear-cut distinction between depressive groups, the authors
concluded that the patients were best considered to form a "continuous
series."

Lewis. At the time of Gillespie's study, and soon after the 1926 meeting,
Lewis was working on his MD. Influenced by Mapother, Lewis carried
out an extensive clinical (catamnestic) study of 61 patients admitted to
the Maudsley Hospital in London. He carefully evaluated a wide range
of clinical features of depression and considered in depth whether there
was a "definite situational factor responsible for the precipitating or the
maintenance of symptoms" (Lewis 1934). He concluded that there was
no clear demarcation between depressive types:
[T]here can be little doubt that this view is false, every illness is the prod-
uct of two factors - of environment working on the organism; whether the
constitutional factor is the predominant determining influence, or the envi-
ronmental one, is never a question of kind, never a question to be dealt with
as an "either/or" problem; there will be a great number of possible combi-
nations according to the individual inherited endowment in training, and the
particular constellation of environmental forces. To set up a sharp distinction
"in the interest of academic accuracy," f when the distinction is not found in
f An allusion to Buzzard.
14 P. Boyce and D. Hadzi-Pavlovic
nature, this is no help to thought or action. ... [I]t is very probable, that all
the tables and classifications and terms of symptoms are nothing more than
attempts to distinguish between acute and chronic, mild and severe; and where
two categories only are presented, the one - manic depressive - gives the char-
acteristics of acute, severe depression, the other of chronic mild depression.
(Lewis 1934)

While Lewis is often considered to have been a strong proponent of


the unitary view of depression - as evidenced by the quote above - this
is not really the case. His position was more pragmatic, more agnostic,
than is often implied. He clearly acknowledged that manic depressive ill-
ness comprises a heterogeneous group of disorders. As he was unable to
clearly distinguish between the depressive types on the basis of his clini-
cal observations, he did not support the endogenous-reactive dichotomy.
But this was not intended to imply support for the alternative, unitary,
view of depression. He recognised the limitations of the approaches
used, particularly the focus on possible aetiological factors (family his-
tory, psychogenesis), and on symptom patterns which, as noted earlier,
often implied degrees of severity. So he preferred to leave the gate open,
since,
No doubt increasing knowledge will bring an improved, eventually even a sta-
ble, classification based on aetiology, and pointing, it may be hoped, to treat-
ment: whether it comes by way of genetics, psychology, or somatic pathology,
it will be welcome so long as it is useful and valid. (Lewis 1938)
In the meantime, his suggested approach to classification was a heuristic
"avoidance of classification":
The classification, or avoidance of classification, which I would propose is:
first, to determine whether the depression is severe or mild - melancholia or
neurasthenia, descriptively speaking; ... What, in sum, is the presumptive
balance of environmental forces responsible for the illness as against inherited
ones; ... It is because we have no sure means of distinguishing exactly the
numerous causes in each case, and their effects, that we must deny ourselves
the ease of a simple classification. (Lewis 1938)
The end of all his research was to return - perhaps never to leave - the
biopsychosocial model of his teacher Meyer.

Why the Debate?


The classification issue, at least in the early years of the debate, centred
around the two major issues of aetiology and symptom pattern. Re-
garding aetiology, the role of psychogenesis was viewed as central: reac-
Some Historical Aspects 15
tive depressions were precipitated and endogenous were not. Regarding
symptom pattern, the question was whether types could be distinguished
on the basis of symptom patterns.

Semantics of Depression: I. Endogeneity. When Kraepelin classi-


fied manic depressive illness as an endogenous illness he used Mobius's
formulation of "endogenous." This starts by distinguishing between (i) a
"main cause" or "conditio sine qua non," and (ii) "subsidiary causes"
(translation by Lewis 1971). Using this distinction, then, endogenous
diseases are those whose main cause is a factor internal to the indi-
vidual, and exogenous diseases are those whose main cause must come
from outside the individual. "Endogenous" had a narrower meaning
than might appear, because, for Mobius, an internal factor meant a
hereditary predisposition associated with a very high probability of the
disease developing.
This definition had a number of features which made it difficult to
apply rigorously to psychiatric illness:
(i) the need to specify a clear hierarchy of causes;
(ii) genetic inevitability not usually being evident in psychiatric disor-
ders;
(iii) the apparent exclusion - as endogenous main causes - of psychic
structures laid down in early childhood; and
(iv) the sense that endogenous = unprecipitated, when unequivocally
unprecipitated primary depressions are difficult to demonstrate con-
vincingly.
One plausible reading of this particular rendering of the endogen-
ous/exogenous dichotomy is that it is neutral about secondary causes
- an endogenous disease can be set in train by an exogenous, sub-
sidiary, cause. Hence, pointing to an endogenous depression preceded
by a "causative" life event does not provide grounds for rejecting the
dichotomy.

Semantics of Depression: II. Reactivity. Gillespie tried to avoid


the above difficulties by shifting the ground. Rather than attend to
aetiology, his differentiation of depressive types was based almost totally
on the course of the illness after, and irrespective of how, it was initiated.
The term "reactive" is used in a wider but more consistent sense than usual.
The reactivity in the sense of this paper is by definition a general one, not
confined to the precipitating circumstances, but occurring also throughout
16 P. Boyce and D. Hadzi-Pavlovic
the course of the depression. ... The central feature of the affective condition
in these patients [reactive depressions] was not merely its variability (which
existed to some extent in nearly every case in every group, but was especially
pronounced in this one), but its responsiveness to influence, both external and
internal (conscious). This responsiveness, by virtue of which one has ventured
to call the group "reactive," was exhibited in relation to a variety of factors.
... The autonomous depressions (i.e. those showing no reactivity beyond at
most an initial precipitation) are a very heterogeneous group. (Gillespie 1929)

For Gillespie the course of the depression did not mean just the pro-
cession of recovery and improvement, but the pattern of symptoms over
time. Illness might be initiated by, and the course determined by, any
combination of internal factors and external factors, but reactivity pro-
vided the best basis for typology.
This use of the term "reactive," however, was not without confu-
sion. Lewis (1934) applied different sets of criteria for reactivity to his
61 patients. Using the first of these (due to Dawson), Lewis rejected
Gillespie's dichotomy because he was able to find only 10 patients in
his 61 for whom no clear-cut precipitating factor to their depression
could be identified - that is, most patients were reactive (in the sense
of precipitated). As a test of Gillespie's division into autonomous and
reactive, however, these criteria placed far too much weight on initial
precipitation. A second set that Lewis applied were Jasper's criteria for
reactivity, the third of which was that "the course of the psychoses must
be dependent on changes in the situation." Lewis commented that this
"third requirement was satisfied in four [patients]," which would imply
that most patients were now non-reactive, that is, autonomous. Overall,
Lewis judged 9/61 as exogenous (as per Mobius) and 10/61 as endog-
enous; the remaining two-thirds only emphasised the sheer difficulty in
applying the criteria - for endogeneity and reactivity - to individual
patients.

Sorting by Symptoms. One of the major reasons that people adopted


the Unitarian position (or rejected the binary position) was the perceived
inability of symptoms to cluster into distinct depressive subgroups. It
must be remembered that not only were these studies conducted well
before computers made a large range of statistical analyses practical,
but the small sample sizes would have limited any available analyses.

Patient Sources. The earliest clinical observations were predominantly


based on hospitalised patients, while those with the milder neurotic de-
Some Historical Aspects 17
pressions and adjustment disorders were rarely seen by psychiatrists.
Another important consideration regarding those hospitalised with de-
pression was that they were often admitted to hospital with a view to
certification. As Ross (1924) put it
A large number of manic depressives never come under the care of those in
charge of mental hospitals, for they are not certifiable because they do not
show any of those disorders of conduct which would make that procedure
necessary or justifiable.

This would mean that people admitted to hospital with depression were
clearly at the more severe end of the spectrum of people suffering from
depressive illness of whatever type. Perhaps an exception were those
persons admitted to the Cassel Hospital, which had a psychoanalyti-
cally based orientation. It is noteworthy that in that hospital Gillespie
was able to identify different types of depression quite clearly. The sit-
uation with the Maudsley Hospital - where Mapother and Lewis found
only continuity - is more difficult, since that hospital was able to ac-
cept voluntary non-certifiable patients and emphasised early interven-
tion. Yet Ross was obviously sceptical about the patients who were seen
by Mapother, and he was summarised (Mapother 1926) as stating,
the fact that Dr. Mapother emphasized neurology, and not cardiology, gas-
trology, proctology, and all the other ologies of the body, made the speaker
suspect that it was only a very small section of the psychoneurotic group that
found its way to the Maudsley Hospital.
To which Mapother replied that
for the past three [years he] had been connected with the only hospital in this
country where it was possible to get simultaneous experience of both extremes
in mental disorder and of the intermediate cases that were the crux of this
question.

Current Status
In many ways both the form of the debate and its ground rules, as
established in the 1920s, are still with us - though the tools and aims
have shifted around.

Through a Computer Darkly. Diagnosis - and, more so, establishing


diagnostic entities - is a difficult task. Complex diseases, with many
forms of presentation, are not easily captured without larger sample
sizes and longitudinal observation. James Parkinson, for example, with
18 P. Boyce and D. Hadzi-Pavlovic
a sample of seven, gave little weight to rigidity in his descriptions of
Parkinson's disease, and it was 50 years before Charcot established it
as a cardinal sign (Jankovic 1987). Parkinson also asserted that there
was no change in the mental functioning of his patients - with dispute
of that assertion continuing for over 100 years (Goetz 1992). A major
extension of the debate came with the easy applicability of multivariate
techniques afforded by computers. Clinicians' gestalt, and the simple
tallying of symptoms and signs, were cast aside for factor loadings and
the distributions of discriminant function scores. Wise men were guided
by stars of significance rather than constellations of symptoms, and the
debate became a mathematical, as much as a clinical, debate (such as
the thread starting with Moran [1966], continuing with Maxwell [1971]
and Eysenck [1970] and then followed up by Grayson [1987]).

Prescription Trumps Description. For a long time it was possible to


dismiss classification as the mere academic shuffling of suspect entities.
This changed in the 1950s as, to the already extant electroconvulsive
therapy (ECT), were added lithium and the more broadly applicable,
monoamine oxidase inhibitors and tricyclic antidepressants. Initially,
the existence of treatments that were both effective and biologically
focussed was seen to raise questions not only about who would benefit,
but also about the true nature of depression. This view could be said to
have had its official apogee in one of the DSM-III-R melancholia criteria -
previous response to antidepressants - but in truth it was already a
relic, beached by a tide long gone elsewhere. Well before DSM-III-R, the
widespread dispensing of antidepressants, and a generous perception of
their effectiveness, had created a sense of one disease, one treatment,
one outcome, which seemed to make classification irrelevant once again
(Boyce and Hickie 1994).
We have tried to indicate some of the things which made (and which
make) the debate about classification difficult. In the absence of a knock-
down argument the only sensible verdict would seem to be the Scottish
verdict - not proven - but a view must prevail, and one did. The unitary
view.
Since Meyer, American psychiatry had had a dimensional view. This
biopsycho(dynamic)social model was not one which easily accommo-
dated either mental illness or categories of such an illness. The advances
of biological psychiatry may require researchers to move to categorical
subgroups, homogeneous for certain characteristics.
In English psychiatry, Lewis carried the day. Re-reading him we note
Some Historical Aspects 19
that his conclusions were far less final than history makes them seem.
More importantly, we recognise the need for the proponents of a par-
ticular view to present a clinically convincing case. This book argues
that by re-examining certain clinical features which once were regarded
as crucial - namely psychomotor disturbance - a far stronger statement
of the binary (or non-unitary) position can be made.
2
Issues in Classification:
II. Classifying Melancholia
GORDON PARKER
DUSAN HADZI-PAVLOVIC
PHILIP BOYCE

Introduction
In the previous chapter, an overview was provided on the Unitarian ver-
sus binarian debate. Two issues briefly noted there will be considered in
greater depth in this chapter: firstly, the extent to which certain clinical
features have been demonstrated as either specific to or over-represented
in the melancholic type of depression; secondly, the extent to which the
debate has been assisted by statistical analyses. Both issues converge in
our progressive attempt to define melancholia, and in considering its sta-
tus - as either a discrete categorical entity with specific clinical features
or a dimensionally severe expression of depression.

Clinical Features of Melancholia


While there has clearly been an extended controversy about the sta-
tus and classification of melancholia (synonyms include "endogenous,"
"endogenomorphic" and "autonomous" depression), suggested defining
characteristics (e.g., Kendell 1968; Klerman 1971; Young et al. 1986)
have included:
(i) a distinctive pattern of symptoms and signs;
(ii) the greater relevance of genetic and other biological (as opposed to
psychosocial) determinants;
(iii) concomitant evidence of disturbed biological functioning, especially
of the hypothalamic-pituitary-adrenal axis; and
(iv) selective response to physical treatments (i.e., antidepressant med-
ication and electroconvulsive therapy).
Our focus, both in this section and in this book, is on the clinical defi-
nition of melancholia and the extent to which symptoms and signs either

20
Classifying Melancholia 21
have specificity to or are over-represented in melancholia. In essence, we
respect the view of Ni Brolchain that classification should be restricted
"in the first instance, to mental state items. Having done this, one
may go on to examine whether and how aetiological and background
factors, age, personality traits, etc., are associated with the clinically-
defined classes" (Ni Brolchain 1979). A large list of suggested specific
features can be readily assembled from just a few reviews (e.g., Rosenthal
and Klerman 1966; Nelson and Charney 1981; Rush and Weissenburger
1994), and we group these suggested features of melancholia under sev-
eral headings:
(i) Mood state items
(a) guilt (including "severe" and "pathological")
(b) remorse, self-reproach, self-pity
(c) unworthiness, hopelessness
(d) severity of mood
(e) non-reactive mood
(f) loss of interest
(g) anhedonia (general, anticipatory and consummatory)
(h) "distinct quality"
(i) suicidal ideation or attempts
(ii) "Vegetative" items
(a) loss of appetite and/or weight loss
(b) insomnia (initial, middle and, in particular, terminal)
(iii) Diurnal variation
(a) mood and/or energy worse in the morning
(iv) Other
(a) retardation
(b) agitation
(c) concentration difficulties
(d) psychotic features (i.e., delusions, hallucinations)
Such a lengthy and comprehensive list might well appear, at face
value, to define most of the parameters of depression per se, rather than
any melancholic type. This, of course, would not be inappropriate if
melancholia is a more severe expression of depression, since it would then
be expected that all clinical features of depression would be represented.
As noted in the next section, there have been numerous attempts
to refine that lengthy list of clinical features. Rather than consider
22 G. Parker, D. Hadzi-Pavlovic and P. Boyce

the many individual studies, we first note the review by Nelson and
Charney (1981). They examined 33 studies using multivariate analytic
approaches (20 having used factor analysis, 9 cluster analysis, and 4 dis-
criminant function analysis), and their summary is informative. They
concluded that "psychomotor change is the symptom most strongly asso-
ciated," with retardation being somewhat more consistently associated
than agitation. Several other features well supported were: non-reactive
mood, a severely depressed mood, loss of interest in pleasurable activity
(or anhedonia), depressive delusions and self-reproach. Features such as
"distinct quality," early morning awakening and concentration difficul-
ties received only moderate support; while "vegetative" features such as
appetite loss, weight loss and insomnia received little support.
As that review was necessarily impressionistic in estimating consis-
tency across a number of studies of variable sophistication, we attempted
a more quantitative synthesis (Parker, Hadzi-Pavlovic and Boyce 1989),
by examining a number of the more influential factor analytic studies to
assess the extent to which they had identified an "endogenous" set of
features with consistency, and then assessing the relative importance of
features contributing to that set. In selecting studies, we limited con-
sideration (1) to variables which had been examined in at least three
studies; and (2) to studies (a) which sampled patient groups rather
than non-clinical subjects; (b) which restricted consideration to sub-
jects diagnosed as having a primary depressive disorder and (c) where
an "endogenous" factor was identified by the researchers.
First, we examined the levels of congruence and similarity of factor
loadings across those nine studies (Hamilton and White 1959; Kiloh
and Garside 1963; Carney, Roth and Garside 1965; Rosenthal and Kler-
man 1966; Rosenthal and Gudeman 1967; Mendels and Cochrane 1968;
Paykel, Prusoff and Klerman 1971; Garside et al. 1971; Kiloh et al. 1971),
with moderate to high agreement being demonstrated for the majority
of studies. Next we ranked variables in terms of their importance in
contributing to an "endogenous" factor, converting the factor loadings
to (normally distributed) z values using Fisher's r-z transformation and
summing across the studies. In relation to clinical features, the high-
est mean score ranking was severity, followed in turn by retardation,
non-reactivity, distinct quality, non-variable mood, delusions/paranoid
features and guilt. Lower-ranking features included terminal insomnia,
agitation, diurnal variation and weight loss.
A more recent review also provides information on refined features.
In examining the clinical utility of preserving melancholia in the DSM-
Classifying Melancholia 23
IV system, Rush and Weissenburger (1994) examined nine "diagnostic
operationalizations of the melancholic (endogenous) concept" (including
DSM-m, DSM-III-R, two versions of the Newcastle Scale, and the Hamil-
ton Depression Rating Scale). The only criterion or feature included
in all systems was psychomotor retardation. Terminal insomnia, early
morning awakening and weight loss were represented in seven; psychomo-
tor agitation in six; guilt in five; distinct quality, delusions, anhedonia
and appetite loss in four; and non-reactive mood and loss of interest in
three. Minimally represented were middle insomnia; hopelessness, sui-
cidal thoughts or behaviour; and sad, anxious or dysphoric mood. In
addition to noting that melancholic features occur in nearly all those
with psychotic depression, Rush and Weissenburger said that their re-
view indicated that such melancholic features predicted a good response
to electroconvulsive therapy (ECT) and may predict a good response to
antidepressants. Their summary (considering clinical features) is worth
noting: "Melancholic features included most consistently in clinical diag-
nostic systems are psychomotor retardation and agitation, late insomnia,
weight and appetite loss, anhedonia, distinct quality of mood, diurnal
variation, guilt and delusions."
Finally, in this section, we note a study of clinicians' diagnostic habits.
If a diagnosis (melancholia or other) has accepted criteria, then such a
study is pointless, as the diagnostic criteria will be affirmed, whether
they validly define clinical features of melancholia or not. While our
study (Parker et al. 1989) involved 27 psychiatrists generating DSM-III,
ICD-9 and Mood Disorders Unit (MDU) clinical diagnoses on 300 de-
pressed patients, we suspect that our reliance on clinical psychiatrists
resulted in a tendency to rate according to their own clinical rules -
and so allowed us to obtain a clinical world view on features generating
diagnostic decisions. We had two potential melancholic (i.e., psychotic
and endogenous depression) and two non-melancholic (i.e., neurotic and
reactive depression) classes. In terms of mental state signs, the com-
bined psychotic/endogenous patients rated significantly higher than the
combined neurotic/reactive depressives most clearly on items suggest-
ing: psychomotor retardation (poverty of associations, immobile face,
dull/inattentive, immobility, non-reactivity [i.e., non-responsive to in-
terviewer; unable to be cheered up]); guilt and nihilism; and a variable
termed "endogenous quality" - although we unfortunately failed to ask
raters to define the qualitative components that influenced their ratings.
When the previously aggregated psychotic and endogenous depressives
were compared with each other on mental state signs, the diagnosed
24 G. Parker, D. Hadzi-Pavlovic and P. Boyce
psychotic depressives tended to rate as both more retarded and agitated
than the endogenous depressives.
We then examined symptom-based data. Some features (sadness, an-
ergia, poor concentration, slowed thoughts, anhedonia and reduced abil-
ity to work) were rated as present in the great majority of subjects in
each broad diagnostic group and therefore had minimal discriminatory
potential. When the psychotic/endogenous depressives were contrasted
with the neurotic/reactive group, the former were most likely to report
mood non-reactivity. They were also more likely to report delusions and
hallucinations, to have greater difficulty working, to be constipated and
to judge their mood as non-varying. Findings in regard to weight change
were somewhat unclear. Against expectation, no significant differences
were established for certain symptoms, including guilt, a sense of deserv-
ing punishment, insomnia (including early morning wakening), diurnal
variation of mood and energy, suicidal thoughts, subjective feelings of
retardation and agitation, and appetite or eating changes. When the
previously aggregated psychotic and endogenous depressives were com-
pared with each other, only a few differences in ratings were established.
The psychotic depressives were more likely to have delusions and hallu-
cinations (by definition) and to affirm a sense of deserving punishment,
were less likely to report a diurnal pattern in energy and were more
likely to report constipation, loss of anticipatory pleasure, inability to
be cheered up and rapid weight loss. The suggested differential relevance
of features (e.g., psychotic disturbance) when assessed as symptoms, as
against signs, encouraged our research focus on mental state signs of
melancholia.

Discussion: Clinical Features of Melancholia


In considering here whether certain clinical features have specificity to
melancholia, we have adopted a broad "convergence of evidence" ap-
proach. As we do not know whether melancholia is a categorical entity
(in and of itself, or encompassing psychotic depression as well), and as
there is no gold standard against which we can examine suggested fea-
tures, precise conclusions cannot be reached and there would be little
value in considering individual studies in detail. We have therefore ap-
proached the issue by examining for any consistency in identification
of melancholic items across somewhat differing strategies: first, by con-
sidering reviews of published multivariate studies; secondly, by examin-
ing an empirical quantification of congruity across multivariate studies;
Classifying Melancholia 25
and, thirdly, by examining clinical "fashion" or diagnostic "habit." Even
if consistency across studies is demonstrated, validity cannot necessar-
ily be claimed. For instance, if there is a key clinical feature defining
melancholia but which has either never or rarely been included in data
sets (or alternately never been measured appropriately), it is unlikely
to meet any "consistency" criterion. Conversely, consistency may be
demonstrated (a) in the absence of any specificity to the melancholic
"type" (for variables that are ubiquitous across depressive types); and
(b) as a consequence of a reified false assumption or interpretation. As
explanation of the latter, let us imagine that a diagnostic system incor-
rectly includes an invalid variable (say poor concentration) as a criterion
for "melancholia" - invalid in this instance by having no specificity to
melancholia, however much it might be reported by those with or with-
out melancholia. An item on concentration might then appear in all
melancholia indices, with clinicians being encouraged to rate all those
with poor concentration as having "melancholia," and with the concen-
tration item then having some spurious specificity to melancholia. By
such processes, an invalid item may pass our several assessments of con-
sistency.
After respecting such caveats we suggest that the several assessment
strategies indicate varying levels of probability for a number of items
as having specificity to melancholia. Across all assessment approaches,
psychomotor disturbance is both the most consistently suggested and
most discriminating feature, especially when measured as an observed
sign. (As a symptom, however, "feeling slowed down" tends to be a
rather ubiquitous depressive symptom rather than having specificity to
melancholia.) When disaggregated into psychomotor retardation and
agitation components, retardation remains clearly indicated, while agi-
tation is variably suggested as strongly or minimally specific. This may
well be a reflection of measurement, with loss of specificity when some
systems may allow (by definition or default) for any physical symptoms
or signs of "anxiety" to be interpreted as "agitation," diluting any speci-
ficity emerging from "true" agitation. Other consistently suggested fea-
tures are (a) severity of mood disturbance (and perhaps some depressed
mood concomitants such as loss of interest, and self-reproach); and (b) a
non-reactive (or, as sometimes described, a non-variable) mood. While
not consistent, because psychotically depressed patients are variably in-
cluded or excluded in the definition of melancholia, psychotic features
have high specificity (although often a low prevalence) in the melancholic
class. Such features include delusions and hallucinations, while others
26 G. Parker, D. Hadzi-Pavlovic and P. Boyce
(such as guilt and nihilism) are variably suggested - perhaps achieving
specificity when positive ratings reflect representation at delusional or
over-valued idea levels.
A number of items remain problematic, perhaps because definitional
difficulties produced questionable specificity. "Distinct quality" had
quite varying utility across our several strategies - presumably because
of its indistinct meaning. Distinct quality was defined in the DSM-III
(American Psychiatric Association 1980) criteria set for melancholia as
follows: "The depressed mood is perceived as distinctly different from
the kind of feeling experienced following the death of a loved one" and,
while deleted as a DSM-III-R (American Psychiatric Association 1987)
criterion, has returned in the DSM-IV (American Psychiatric Association
1994) system. The problem with such a definition is that it fails to
offer any positive defining characteristics. The death of a loved one is
generally viewed as a grief experience, and grief is commonly seen phe-
nomenologically (e.g., Parkes 1973) as akin to separation anxiety rather
than to depression. Thus, if we exclude such anxiety-based experiences,
what is left for the individual to report as feeling "distinctly different" ?
Answer: conceivably all depressive experiences, so that distinct quality
may mean no more than feeling depressed. Guilt is generally indistinctly
defined. The relevant item in the DSM-IV criteria set for major depressive
episode is "feelings of worthlessness or excessive or inappropriate guilt
(which may be delusional) nearly every day (not merely self-reproach or
guilt about being sick)." Such a definition allows positive ratings to be
returned for guilt ranging from a relatively mild expression through to
a psychotic level, yet it may be that guilt only achieves specificity to
melancholia when it is at the level of an over-valued idea or delusion.
The so-called vegetative features (e.g., diurnal variation in mood, ap-
petite and weight loss, terminal insomnia) were variably suggested as
either quite strongly or as minimally differentiating. While common in
melancholia, such features may also be common in non-melancholic de-
pression - or at least certain non-melancholic sub-groups. Thus, those
who have an acute reactive depression (akin to a grief process) commonly
report severe insomnia, while the severity of appetite and weight loss in
such sub-groups may exceed that generally observed in melancholia.
The last point indicates other problems contributing to any considera-
tion of specificity. Should specificity refer only to differential prevalence
and/or refer to severity? How valid are symptom reports, especially
those of melancholic and psychotic depressive patients? Should features
be rated at the nadir of episode or at presentation (when an individ-
Classifying Melancholia 27
ual may lose weight initially, regain weight and then, following self-
consolatory strategies or medication effects, gain weight)? To what ex-
tent do medications (especially antidepressant and antipsychotic drugs)
have artifactual influences on reporting? Clearly, there is a need for
lucid operational definitions of a number of items, with optimal rating
strategies being identified before clear-cut conclusions about specificity
are allowed. In our empirical studies reviewed in subsequent chapters,
we have therefore been liberal in including a wide set of potential items
initially, before proceeding to refine that list.

Limitations to Previous Analytic Strategies


The binary school argues that there are at least two types of depression
which differ in their symptoms, aetiology and outcome. The Unitarians
postulate that the data are better represented by a single depressive
type which varies dimensionally. The debate between supporters of the
unitary and binary views moved from clinical intuition to interpretation
of empirical analyses in the sixties, with the binary view subsequently
being argued principally on the basis of two types of analyses (factor
analytic methods and examination for evidence of bimodality), and with
limitations to both those analyses rarely being appreciated. We shall
consider concerns about each in turn.
Until recently, the strongest statistical support for the binary view
of depression has been provided by factor (principal components) an-
alytic studies which generally delineated a bipolar factor with features
interpreted as reflecting endogenous depression and neurotic depression
at opposing poles. A review of this literature, however, suggests to us
that previous studies may not have separated endogenous and neurotic
depressive entities as claimed. If our re-interpretation is correct, the
features claimed to describe and contrast depressive types may not only
be in error, but have led to significant limitations in depression research
and clinical practice.

Factor Analytic Approaches


Factor analytic methods are not really capable of delivering a typol-
ogy (in that they produce dimensions of symptoms, not groupings of
patients). In the same way that a class of students may be divided
into sub-groups or sub-classes (e.g., males vs. females) or along dimen-
sions (e.g., height, intelligence), resolution of the unitary/binary debate
28 G. Parker, D. Hadzi-Pavlovic and P. Boyce
requires an analytic approach capable of distinguishing a melancholic
class from a residual non-melancholic class. Factor analytic approaches
can, at best, do no more than provide an indirect way of examining ty-
pological hypotheses by describing important dimensions of variation -
in grouping together symptom patterns among individuals who attract
a diagnosis of depression.
However, as the bulk of the multivariate studies available to the bi-
narian and Unitarian protagonists have been factor or principal compo-
nents analyses (FA and PCA) of depressive symptoms, there is benefit
in detailing some other limitations to their application. We therefore
provide an overview of key studies, demonstrate the relatively consis-
tent interpretation of a bipolar factor as one contrasting "endogenous"
with "neurotic/non-endogenous" depressive features, and then argue
that most of those interpretations are capable of a distinctly differing in-
terpretation. Our overview is derived from a detailed earlier publication
(Parker et al. 1991c).
Eysenck (1970) argued that the unitary/binary debate could be "trans-
lated into factor analytic logic quite simply and clearly." Specifically, the
unitary view predicts that a factor analysis of depressive symptoms will
produce a single large general factor with positive loadings and the ab-
sence of any interpret able bipolar factor. The binary view, by contrast,
predicts that there will be a bipolar factor contrasting endogenous and
non-endogenous symptoms in addition to the general factor. Eysenck
argued, "All the protagonists have found a clear-cut positive answer ...
the unitary hypothesis is wrong, and the binary hypothesis is supported
very strongly indeed."
Our first caveat is that the mere reporting of a two-factor solution
does not, however, establish the existence of two separate diagnostic
categories, be they categorical or dimensional, particularly when the
researchers have generally extracted, or interpreted, only the first two
factors. The binary view is not supported by the mere presence of a
bipolar factor, for bipolarity is a necessary "property of the arithmetic
procedure of factor analysis" (Maxwell 1971). If the sample contains
only homogeneous groups of melancholic and non-melancholic depres-
sives, then one of the relevant dimensions should contrast the so-called
(according to the binarian terminology) "endogenous" and "neurotic"
features of depression. If the sample contains higher-order groupings
(e.g., those with and without primary depression), as we will argue here
have occurred in the majority of studies, the derived bipolar factor will
tend to contrast respective symptoms of the "depressives" and of the
Classifying Melancholia 29
"others." Instead, it is the emergence of a particular and interpretable
factor that is of relevance (Maxwell 1971; Moran 1966), namely, one that
consistently identifies a unique and coherent set of clinical features of
one or more depressive types which both corresponds to clinical expe-
rience, and which may be validated against aetiological factors, course
of illness and treatment response. In the light of these provisos, we will
argue that the use of factor analytic methods has not settled the issue
of what characterises the contrasting features in the bipolar factor la-
belled "endogenous-neurotic" in most of the studies, and has restricted
definition of truly specific clinical features of "melancholia."
In Table 2.1 we consider a number of FA or PC A studies which have
identified a bipolar factor (be it the first or second factor derived), and
list the items making a significant loading, presenting in italics those
which we interpret as suggestive of anxiety and in bold those items sug-
gestive of atypical personality. In the first study, Hamilton and White
(1959) interpreted their findings as supporting the clinical view that "en-
dogenous depression tends to be of the retarded type." In a later PC A
study, Hamilton (1967) interpreted the bipolar factor as representing the
"traditional clinical dichotomy of Retarded vs. Agitated Depression."
We suggest that in these earliest studies the bipolar factor is more par-
simoniously viewed as contrasting no more than expressions of anxiety
and depression.
In the first of the "Newcastle Studies," Kiloh and Garside (1963) re-
ported a seminal study of 143 depressed patients, submitting 35 items to
a simple summation FA and extracting two factors. The first factor had
predominantly positive loadings (Table 2.1), and was labelled a "gen-
eral factor" reflecting "depressive illness as defined by the sum of the
35 features"; we agree with the general nature of the first factor. The
second factor was bipolar, and Table 2.1 again lists the highest loading
features. Because the negatively and positively loading items correlated
significantly with clinical diagnoses of endogenous depression and neu-
rotic depression respectively, the authors concluded that the "bipolar
second factor differentiates between neurotic and endogenous depres-
sion," and that the "data cannot be produced by a single depressive
condition, but must be produced by two separate conditions," viz. en-
dogenous and neurotic depression as two symptom states. We suggest,
however, that the positive loading items define vulnerable personality
characteristics, as well as reactivity and sensitivity to stress, with few
actual depressive symptoms at that "neurotic depression" pole, and pre-
sumably over-represented instead by patients with primary personality
30 G. Parker, D. Hadzi-Pavlovic and P. Boyce

Table 2.1. Factor loadings in representative factor analytic (or prin-


cipal components) studies, contrasting "endogenous" and "neurotic"
depressives, with anxiety and personality features highlighted

Study Factor Pole High loading items


Hamilton (1960) I + Depressed mood, guilt, retardation, loss of in-
sight, suicidal ideation, loss of libido, loss of in-
terest, weight loss, delayed insomnia
— Somatic anxiety, psychic anxiety
II -f Insomnia, gastrointestinal symptoms, agitation,
weight loss, hypochondriasis, psychic anxiety
— Retardation, suicidal thoughts
Hamilton (1967) II + Psychic anxiety, somatic anxiety, agitation
— Retardation
Kiloh&Garside I + Failure of concentration, agitation, suicidal
(1963) ideas, reactive depressive personality, self-
reproach/guilt, hysterical features /imm-
aturity
II + Reactivity, evident precipitant, hysterical, in-
adequate and immature features, vari-
able course, self-pity, reactive depressive
personality
Early awakening, retardation, depression worse
in the morning, distinct quality, greater depth,
short duration, weight loss
Kiloh et al. II -fPrecipitation, reactivity of depression, variabil-
(1972) ity, reactive depressive personality, irri-
tability, inadequate personality, hysterical
personality, hysterical features
I — Obsessional personality, retardation, early
wakening, distinct quality, depression worse in
the morning, agitation
Carney et al. I + Delusions, depressive hallucinations, depressive
(1965) psychomotor activity, "varying" depression, no
reactivity, distinct quality
— Anxiety, hopeful outlook, hypochondriacal, du-
ration > 1 year, self-pity, irritable, blames oth-
ers, phobias
Paykel et al. II + Neuroticism, irritability, initial insomnia, de-
(1971) personalisation, somatic anxiety, suicidal ten-
dencies, obsessional symptoms
— Older age, loss of insight, retardation, diurnal
variation, worse in the morning, distinct qual-
ity, severity
Items suggestive of anxiety are in italics, while items suggestive of atypical
personality are in bold. Reprinted from Parker et al. 1991c.
Classifying Melancholia 31
and anxiety symptom diagnoses. That interpretation is supported by a
description of the subjects in an earlier paper (Kiloh, Ball and Garside
1962), where it is stated that clinical diagnoses were made, in part, by
the presence or absence of neurotic personality traits respectively.
Table 2.1 lists a number of other representative studies, allowing us
to suggest, as has Mullaney (1984) after a review of six factor analytic
studies, that the so-called "endogenous vs. neurotic depression" bipolar
factor is worthy of reinterpretation, with the endogenous pole not being
a constellation of endogenous depressive symptoms alone but perhaps an
admixture of such symptoms as well as symptoms of depression per se,
while the neurotic pole reflects a variable set of features embracing anx-
iety and/or a vulnerable personality style (and very few clinical features
of depression at all).
Our reinterpretation of the endogenous versus neurotic depressive
bipolar factor would be supported by the demonstration of a bipolar
factor in mixed samples of anxious and depressed patients similar to
that identified in depressed samples, and, secondly, its disappearance if
anxiety items are excluded from the data set. We examine the first issue
by reference to a study of Roth et al. (1972) which reported a PCA of 58
variables from in-patients diagnosed as having an anxiety state (n = 68),
a depressive illness (n = 62), or as belonging to a doubtful category
(n = 15). The first factor was a bipolar one with anxiety/neurotic vari-
ables loading positively and depressive features loading negatively. The
authors noted that "most of the maladaptive and neurotic premorbid
personality traits are in ... association with anxiety symptoms rather
than with depressive symptoms." Since scores on derived anxiety and
depressive components correlated negatively (—0.52), they noted that
"the presence of one of the syndromes lessens the chance that the other
is present." They also observed that their bipolar factor resembled a
factor described in a number of earlier FA and PCA studies of depressed
patients only. Additionally, they acknowledged that "maladaptive and
neurotic traits" had been shown in a number of studies to "be con-
centrated at the opposite pole to the endogenous depressive symptom
cluster."
In Table 2.1 we list a group of clinical variables examined in both the
Roth et al. (1972) study of separate anxious and depressed patients and
the Kiloh and Garside (1963) study of depressed patients only, together
with their loadings on the derived bipolar factors. The factor loadings
suggest a strikingly similar pattern, despite using supposedly clinically
distinct diagnostic groups. Roth et al. implicitly foreshadowed our sus-
32 G. Parker, D. Hadzi-Pavlovic and P. Boyce

Table 2.2. Factor loadings generated for comparable variables in studies


of depressives only (Kiloh and Garside 1963) and separate sub-groups
of anxiety and depressive patients (Roth et al. 1972)

Kiloh and Garside study Roth et al. study


Factor II Factor I
Variable Loading Variable Loading
Negative pole
Early awakening -0.69 Early awakening -0.54
Retardation -0.55 Retardation -0.49
Depression worse AM -0.53 Depression worse AM -0.41
Suicide attempts -0.23
Depth of depression -0.37 Depressed mood: severity -0.18
Suicidal ideas -0.13 Suicidal tendencies -0.11
Agitation -0.17 Agitation -0.09
Positive pole
Sudden onset +0.18 Sudden onset +0.13
Hypochondriasis +0.12 Hypochondriacal ideas +0.16
Obsessionality +0.20 Obsessional traits +0.11
Obsessional symptoms +0.46
Anxiety +0.16 Anxiety traits +0.55
Tension +0.15
Panic attacks +0.60
Anxiety symptoms +0.41
Phobias +0.17 Situational phobias +0.60
(agoraphobias)
Other phobias +0.40
Precipitation +0.48 Psychological stress +0.28
Hysterical features +0.37 Hysterical traits +0.47
Hysterical symptoms +0.42
Hysterical features +0.41 Immaturity +0.55
(immaturity)
Variability of illness +0.39 Persistent depression -0.49
Reactivity of depression +0.60 Reactivity of depression +0.40
Reprinted from Parker et al. 1991c.

picion concerning the mislabelling of "neurotic depression" in pointing


out that, unless there are specific criteria for excluding anxiety disorders
from samples of depressed patients, anxiety disorders are "liable to be
represented in any clinical material of depression."
We have provided (Parker et al. 1991c) additional support for the
second point of our argument by reporting two PCA studies (involv-
ing independent samples), where we first included depressive features
suggested as discriminating between endogenous and neurotic/reactive
Classifying Melancholia 33
depressive disorders by Rosenthal and Klerman (1966), and then re-
peated the PCAs after deleting a priori anxiety items. In both studies,
we demonstrated that the neurotic depressive pole could be made to
appear and disappear by the inclusion and exclusion of anxiety items.
When the historically described endogenous versus neurotic depressive
bipolar factor was no longer able to contrast depression and anxiety, it
contrasted lower-order symptom complexes (i.e., retardation vs. vege-
tative symptom complexes in Study I and dichotomised depression vs.
irritation/frustration in Study II).

Discussion of Factor Analytic Studies


We have reconsidered the claim that FA studies have delineated two de-
pressive entities, endogenous and neurotic depression, and put a case for
qualification, challenging the conventional interpretation of a bipolar fac-
tor necessarily contrasting endogenous and neurotic depressive features.
If the revisionist view is accepted, then the features held to define both
depressive entities (endogenous and neurotic depression according to the
binarians) in such studies should not be accepted. Turning to the neu-
rotic depressive pole, we argue that FA studies have not clearly defined
a discrete symptom dimension of neurotic depression, but have instead
mislabelled conglomerates of state and trait characteristics, reflecting
anxiety, personality and other background variables rather than depres-
sive symptoms. The endogenous depression pole, as denned in such
studies, is then likely to contain melancholia-specific clinical features
but, regrettably, clinical features also defining depression per se. Thus,
to the extent that melancholic features are loaded on that pole, valida-
tion strategies would be expected to show associations (albeit weakened
by inclusion of non-specific depression items) between high scores on
such a dimension, and clinical diagnosis, background risk variables and
positive response to physical treatments, as demonstrated in a number
of studies. For instance, scores on the first factor (a bipolar one) in the
study by Carney, Roth and Garside (1965) correlated with both clin-
ical diagnosis and ECT outcome. While such findings supporting the
conventional articulation of the central features of endogenous depres-
sion or melancholia appear to support the validity of the endogenous
pole, we note again that such items are likely to be a mix of both true
melancholia-specific features and non-specific depression items. For that
reason alone, FA studies have provided limited refinement of the clinical
features of endogenous depression or melancholia, and have almost cer-
34 G. Parker, D. Hadzi-Pavlovic and P. Boyce
tainly led to some spurious items. Thus, when Rosenthal and Klerman
(1966) compared FA solutions with historical views, such as those pro-
posed by Gillespie (1929), they identified a rather lengthy list of features,
namely: non-reactivity, severity, feelings of guilt, remorse and unworthi-
ness; insomnia (especially middle and terminal); diurnal variation (with
mood worse in the morning), retardation, agitation, vegetative symp-
toms, decreased libido and amenorrhoea. However, when Nelson and
Charney (1981) reviewed a number of FA together with cluster analytic
studies (more capable of grouping patients) of depressed patients, a more
restricted list of endogeneity features was suggested: retardation, non-
reactivity, severity, delusional thinking, guilt and early morning awak-
ening, while ones such as vegetative features and diurnal variation were
not strongly supported.
Thus, the intrinsic limitations of factor analysis in disallowing a cat-
egorical entity to be distinguished from a dimensional disorder, as well
as limitations to interpretation of a number of applied studies, may well
have caused the binary argument to founder and the default option (i.e.,
depression as essentially a Unitarian condition and perhaps sub-divided
on a severity basis) to be accepted in classificatory systems (such as
the recent DSM-IV [American Psychiatric Association 1994] and ICD-10
[World Health Organization 1992] systems). We suggest, however, that
other statistical techniques are required to test the validity of a modi-
fied binary hypothesis (i.e., that there is categorical depressive entity -
melancholia - and that it may be distinguished on clinical features from
a heterogeneous residue of non-melancholic depressive expressions). An
overview of some more appropriate statistical approaches is provided in
another chapter (Chapter 4), while exploration of the modified binary
hypothesis forms much of the substance of the following chapters.

Is the Demonstration of Bimodality Important?


Kendell (1989) has emphasised that bimodality or demonstration of a
point of rarity is important in order "to demonstrate the existence of
discrete types of depressive illness." He went on to state that bimodality
has generally not been demonstrated or, when demonstrated, had "failed
to survive cross-validation on a second data set . . . or the claims of the
original investigators could not be confirmed by others."
We must be cautious, however, in accepting the view that bimodality
of distributions of clinical features in depressed groups is necessary to
establish the existence of two types of depression. Bimodality can be pro-
Classifying Melancholia 35
duced artefactually (Eysenck 1970; Grayson 1986), such as by selecting
extreme groups on some continuous variable or items that discriminate
best at the middle of an underlying continuous attribute. Conversely,
the absence of bimodality does not necessarily refute the binary view,
because diagnostic and measurement error can prevent it from being de-
tected (Hope 1969; Murphy 1964). We illustrate one such mechanism by
summarising analyses from a recent paper (Parker and Hadzi-Pavlovic
1993).
As noted in Chapter 1, Sir Aubrey Lewis studied 61 depressives in
considerable detail, and concluded that he could find no qualitative dis-
tinctions between the depressed patients (Lewis 1931) and thus estab-
lished himself as a strong and influential advocate of the unitary view
of depression. Subsequently, Kiloh and Garside (1977) extracted data
on 58 variables from Lewis' published papers, and we used their unmod-
ified codings on a reduced set, with 42 assessing both the features of
depression per se as well as those held to distinguish melancholic and
non-melancholic depressive disorders. We undertook a latent class anal-
ysis and refined the set to 14 items best discriminating the two under-
lying classes identified by a mixture analysis. Figure 2.1 demonstrates
that when we created total scores on all 42 clinical feature items, a uni-
modal distribution of scores was created. However, when total scores
on the refined 14 items were plotted, a distinct bimodal distribution
was demonstrated. The reasons for the two distributions are central
to understanding how Lewis' data can be interpreted in quite different
ways. We suggest that Lewis' data set had 14 differentiating clinical
features, while the remaining 28 (being non-differentiating) were more
likely measuring depression severity. We argue then that many histori-
cal attempts to find evidence of separate depressive types or classes may
have failed because the items defining the types were obscured by the in-
clusion of non-typological items, which, when affirmed, reflect and build
more to an over-riding severity dimension. In addition, we suggest that
this problem is carried through in many current melancholia indices and
diagnostic criteria sets.
If it is to be argued that melancholia is an entity, there is a need to
demonstrate a clinical "point of rarity" from non-melancholic forms of
depression, and strategies such as mixture analyses (and our own em-
pirical data) will be considered in other chapters. The point we seek to
make here is that, while failure to demonstrate bimodality in a number
of data sets (e.g., Kendell 1968) does allow the researcher to conclude
the analysis "fails ... to provide any evidence of a qualitative difference"
36 G. Parker, D. Hadzi-Pavlovic and P. Boyce

I!

4 6 8 10 6 8 10 12 14 16 18 20 22
(a) Total score on 14 items (b) Total score on 42 items

Fig. 2.1. Frequency Distributions for (a) Total Scores on 14 Refined Items
and (b) on 42 Clinical Features. The normal distributions fitted using mixture
analysis are represented by the smoothed curves. Adapted from Parker and
Hadzi-Pavlovic (1993).

between depressive types (Kendell 1968), it does not of necessity allow


rejection of the binary hypothesis. Clearly, analyses may fail to confirm
or reject hypotheses, and it is important to consider where the failure
resides - be it with the hypothesis or the test procedure. If severity
and sub-typing are independent to some degree (as we would argue for
depression), then the two should not be confounded in the diagnostic
process (or in interpreting plotted scores of depressive features). Di-
agnostic decisions should be predicated on the presence or absence of
specific defining features (which allow, depending on the number of cri-
teria met, estimates of probability of category). They should not be
made by summing clinical features of depression per se, for the latter
Classifying Melancholia 37
dimensional model will confound sub-typing decisions by the imposition
of a severity dimension.

Conclusions
While the list of melancholic or "endogeneity" features has been refined
over time, and certain features identified with some consistency, we have
noted a number of limitations to the identifying strategies and even
to accepting items on the basis of consistency. As melancholia may
both have specific clinical features and, overall, be more severe than
non-melancholic depression, there is a need to ensure that any severity-
based influence is addressed in attempting to define melancholia-specific
features. Again, we note dissonance between interpretations based on
symptom versus observational data. These several limitations may well
have contributed to any failure to identify a refined list of clinical features
of melancholia and thus failure to resolve the unitary/binary debate.
3
Issues in Classification:
III. Utilising Behavioural Constructs
in Melancholia Research
IAN HICKIE

Introduction
Much of the work described in this book represents a specific attempt to
identify a key behavioural construct, namely psychomotor change, which
appears to underpin an important clinical entity, namely melancholia.
Conceptually, this represents a different approach to much current clini-
cal research which utilizes broad, non-specific diagnostic categories. The
logical developments of this departure - identification of key constructs,
and the systematic investigation of the patterns of association between
these key constructs - are described. Such patterns of association are
hypothesised to be sensitive indicators of underlying neurobiological pro-
cesses and, hence, the basis of more focussed neurobiological research.
Even if major discoveries occur outside the clinical realm (as is likely in
the realms of molecular biology, genetics and neurochemistry), the ex-
amination of the relationships between such fundamental processes and
refined behavioural constructs will be more useful than studying the
relationships between such processes and broader clinical syndromes.
The application of this approach to current neurobiological research in
melancholia will now be detailed.

Clinical Heterogeneity Impedes Neurobiological Research


Attempts to link broad diagnostic groups in psychiatry with specific neu-
robiological processes are hampered currently by unacceptable hetero-
geneity among formally categorised patient samples. This heterogeneity
is intrinsic to the current ICD-10 (World Health Organization 1992) and
DSM-IV (American Psychiatric Association 1994) international classifi-
cation systems. In an attempt to satisfy clinical, administrative and

38
Behavioural Constructs in Melancholia Research 39
epidemiological needs, these systems have adopted broad class notions
and non-specific diagnostic algorithms. Diagnoses now rely heavily on
self-report symptoms rather than observed clinical signs. Such systems
have accepted the notion of multiple diagnoses, now termed "psychiatric
co-morbidity" (Kessler et al. 1994). Co-morbidity is, however, often
a consequence of overlapping criteria and poor definition of syndrome
boundaries rather than true co-occurrence of independent disorders.
The multiple inter-relationships (at the levels of co-morbidity, lon-
gitudinal history, treatment response and family history) between the
DSM-generated notions of major depression, panic disorder and gener-
alized anxiety illustrate this point (Boyd et al. 1984; Andrews et al.
1990b; Merikangas, Risch and Weissman 1994). Further, clinical phe-
nomenology has been over-simplified to permit the collection of data
by non-clinical research staff in epidemiological settings (Robins et al.
1984; Robins et al. 1988). Key clinical distinctions such as the differ-
ences between recurrent thoughts, obsessions and delusions may be lost
when subjects are interviewed by non-clinical staff in community settings
(Regier et al. 1984; Johnson, Horwath and Weissman 1991).
While the broad DSM approach has significantly advanced reliability
of psychiatric diagnoses, psychiatric epidemiology, service provision and
acceptance of psychiatric disorders in other medical settings, potential
deleterious effects on neurobiological research have been largely ignored.
These adverse effects are well demonstrated in studies of patients with
"major depression" where research has largely failed to:
(i) demonstrate any coherent pattern of neurobiological changes
(Maes et al. 1991; Maes, Calabrese and Meltzer 1994);
(ii) replicate key biological correlates across different research groups,
age cohorts and treatment settings (Carroll 1991; Staner, Linkowski
and Mendlewicz 1994); and
(iii) demonstrate any specific pattern of treatment response outside in-
patient treatment settings (Elkin et al. 1989).
The diagnosis of major depression relies on highly prevalent and non-
specific symptoms, and views self-reported depressed mood as the unify-
ing construct (Robins et al. 1984; Blazer et al. 1994; Blazer et al. 1988).
Such mood disturbances - especially of the limited severity and duration
required for caseness - are extremely prevalent in medical and psychi-
atric practice (Robins et al. 1984; Blazer et al. 1988; Blazer et al. 1994;
Barrett et al. 1988). Consequently, most patient cohorts are highly het-
erogeneous clinically, and the neurobiological characteristics of patients
40 /. Hickie

with "major depression" are then as likely to be correlates of other co-


hort factors (e.g., age, gender, age of onset, duration of illness, treatment
setting, co-morbid anxiety and/or substance abuse) as they are of a pri-
mary mood disorder. As Carroll (1989, 1991) has argued, any proposed
neurobiological maker can only perform as well as the diagnostic sys-
tem with which it is being compared. He demonstrated elegantly that
variability in the available classification systems produced comparable
variability in the performance of the dexamethasone suppression test, a
proposed laboratory marker of melancholia (Carroll et al. 1981a).
Heterogeneity among clinical samples is further complicated by the
diversity of investigators currently involved in neuropsychiatric research
and the range of treatment services from which research samples are
drawn. Individual investigators tend to reflect their own historical tra-
ditions and current practice perspectives. Consequently, concepts of de-
pression derived from clinical psychiatry differ considerably from those
developed in neuropsychology, cognitive psychology, epidemiology and
the basic neurosciences. Patients with major depression recruited from a
private university counselling service or an untreated community sample
share few demographic or clinical features with elderly patients recruited
from tertiary referral, psychogeriatric services.

Reducing Clinical Heterogeneity


In the face of such difficulties, most neurobiological investigators have
resorted to their own case-selection techniques. For example, those con-
cerned with neuroimaging have focused on non-phenomenological (e.g.,
older age, later age of onset, inpatient status and concurrent risk fac-
tors to cerebrovascular disease) and/or clinical features (e.g., presence
of psychotic features or current treatment with ECT) to reduce this
unacceptable heterogeneity (Krishnan et al. 1988; Coffey et al. 1988).
When investigators utilize non-clinical constructs to improve homogene-
ity of patient samples, replication remains problematic. Results derived
may provide greater insights into non-clinical variables (e.g., the neuro-
biological correlates of ageing) than the pathophysiology of the disorder
under investigation. Further, the clinical comparability (e.g., percentage
of psychotic patients, extent of medical co-morbidity) between samples
remains largely inferred, rather than openly stated and/or investigated.
Behavioural Constructs in Melancholia Research 41

Recording Clinical Constructs, Not Diagnoses


Even when attempts are made to collect homogeneous groups (e.g., by
sampling patients over 60 years of age presenting for ECT), patients
still present a wide variety of clinical phenomena. Patients with ma-
jor depression may present with disturbances in one or more of the
following domains: self-reported mood, vegetative symptoms, observed
psychomotor function, observed neurocognitive performance, psychotic
symptoms, interpersonal style, cognitive attributes, subjective energy
and self-destructive behaviour. Additionally, they may vary consider-
ably with regard to critical lifetime (e.g., age of onset, bipolarity, previ-
ous treatments); medical (e.g., smoking, hypertension, thyroid disease);
and family (e.g., history of affective disorder, psychosis and substance
abuse) factors.
Reduction of clinical heterogeneity relies on two factors. First, samples
need to be recruited with respect to the hypothesis under investigation.
For example, studies investigating the pathophysiological correlates of
late-onset depressive disorders should recruit patient and control groups
matched for age, hospitalization and medical status, and then should
restrict their interpretations to psychogeriatric practice with severely ill
patients. Structural brain abnormalities in these patients are unlikely
to be relevant to the majority of patients who develop major depression
early in adult life (Blazer et al. 1988; Blazer et al. 1994). Second, once
carefully selected by referral, treatment, demographic or other factors,
the range of behavioural phenomena presented by patients needs to be
carefully measured.
This two-stage approach has generally not been adopted, as inves-
tigators have typically tried to link simple diagnostic groups (or sin-
gle behavioural constructs) with clinical validators and/or individual
neurobiological processes. More sophisticated approaches have included
using at least a triangular and/or iterative approach whereby the do-
main of clinical diagnosis (or behavioural construct) has been compared
and contrasted with proposed independent clinical validators (e.g., fam-
ily history) and, concurrently, with specific laboratory markers (e.g.,
dexamethasone suppression test [DST]). Due to the multiple compar-
isons required, the actual pattern(s) of association between the key be-
havioural constructs and independent clinical validators and/or putative
neurobiological processes are unlikely to be detected by even the keenest
clinical observer (see Figure 3.1). Further, when one considers that not
all these constructs are necessarily dimensional (e.g., psychosis, observed
42 I. Hickie
psychomotor change), the possible patterns of association become even
more complex.
Different behavioural constructs potentially demand different forms of
measurement. Ideally, those constructs which can be separately identi-
fied, and have been shown to be clinically or neurobiologically relevant,
should be recorded separately. For example, a simple categorical deter-
mination of the presence or absence of psychotic features may assist with
sub-sample analyses, but may mislead as to the relationship between
psychomotor change and psychosis. When a categorical determination
of psychosis is utilized, then patients who are non-psychotic may also
clearly have observable psychomotor change. If, however, a dimensional
rating of psychosis is utilized (with over-valued and guilty ideas being
considered less severe forms of psychosis), then it may not be apparent
that psychosis and psychomotor change are independent constructs.
Non-phenomenological factors, such as age and age of onset of disor-
der, require similar close consideration. For example, if both psychosis
and observed psychomotor change are strongly correlated with age, then
a variety of possible inter-relationships exist (e.g., age as a risk factor
to psychosis and/or psychomotor change, as opposed to age having a
pathoplastic effect). For such complex inter-relationships to be evalu-
ated statistically and longitudinally, the data assessing psychosis and
psychomotor change need to be collected in both categorical and dimen-
sional forms.
As described throughout this book, we have found observed psychomo-
tor change to be a valuable construct in melancholia research. Unlike
subjective anergia, it appears to behave as a categorical variable and,
therefore, constitutes a potential marker of independent depressive sub-
types. Whether that independence extends to underlying neurobiolog-
ical processes is the subject of our current research. Careful differen-
tiation of the sub-components of observed psychomotor change (i.e.,
non-interactiveness, psychomotor retardation and psychomotor agita-
tion) and differentiation from self-reported anergia have allowed us to
consider whether separate pathophysiological processes account for these
varying clinical phenomena.

Latent Statistical Techniques Utilising Behavioural


Constructs
Latent statistical analyses of behavioural and neurobiological data sets
have the potential to reveal complex patterns of associations. The vari-
ous levels at which such patterns may be detected will now be described.
Behavioural Constructs in Melancholia Research 43

////////A

Age
Age of onset
Gender
Duration
Family history
Pre-morbid personality
Social context variables,
Treatment response
Hospitalization
Longitudinal course
Other lifetime diagnoses'
Medical morbidity
Behavioural Constructs/Factors
Bipolar course

Fig. 3.1. Possible Combinations of Behavioural Constructs, Independent Clin-


ical Validators and Neurobiological Processes.
44 /. Hickie
First, there are associations simply within a behavioural domain. For
example, we have used latent class techniques with data concerning psy-
chomotor performance to infer that depressive disorders consist of at
least two sub-types (see Chapter 6).
Secondly, such techniques may be used to study the patterns of as-
sociation between behavioural constructs. For example, various combi-
nations of psychomotor retardation, specific cognitive deficits and delu-
sional ideation may be supported by analyses of clinical data, and this
pattern may then be found to be associated with specific neuroimaging
changes (e.g., deep white matter hyperintensities). By contrast, these
same statistical techniques may suggest that some patients are charac-
terised by mood disturbance and self-destructive behaviour alone, and
that such patients have no detectable abnormalities on structural neuro-
imaging. Hence, the clinical differences between two such groups may
correlate with important structural and functional factors, and suggest
differential neurobiological pathways.
Thirdly, such techniques may be used to generate clinical algorithms
which combine behavioural ratings with socio-demographic and/or
course of illness data. Again, the validity of such algorithms may then be
investigated by comparison with neurobiological data. Such algorithms,
generated empirically, are likely to prove more robust neurobiologically
than those generated by expert committees.
Fourthly, it is possible to include the neurobiological variables them-
selves in the statistical process. Although one is not then able to test any
proposed solutions against independent data (though replication may be
sought in independent samples), one may observe unexpected patterns of
association. Such associations may suggest novel aetiological processes.
For example, if subcortical white matter changes, dopaminergic distur-
bances and evidence of systemic autoimmune activation all co-occur in
older depressives - and are also associated with the clinical features
of psychosis, neurocognitive impairment and psychomotor change - it
might then be inferred that immune processes give rise to subcortical
lesions. In some patients, these lesions then form the neuroanatomic sub-
strate for late-onset psychotic depression, with such features being me-
diated principally within subcortical structures by dopaminergic mech-
anisms.
As the neurocognitive and behavioural constructs described are not
restricted to patients with depression, comparisons of these constructs
in patients with other neuropsychiatric and neurological diagnoses is the
last potential step in this statistical hierarchy. For depression, one needs
Behavioural Constructs in Melancholia Research 45
to compare the statistical patterns (notably for psychomotor change,
mood disturbance and cognitive impairment) with the patterns seen in
patients with those psychiatric (e.g., non-affective psychoses) and neuro-
logical (Parkinson's disease, multiple sclerosis and subcortical dementia)
disorders which result in similar neurocognitive and behavioural abnor-
malities.
Although a wide range of potential predictions can be made, the ex-
tent of data collection will be restricted by patient tolerance, access to
relevant technology and availability of research funds. For melancho-
lia, it can be predicted that the behavioural constructs of psychomotor
change, psychosis and neurocognitive impairment will be strongly associ-
ated and, in turn, that these will be associated with the clinical variables
of late age of onset, and risk factors to cerebrovascular disease. All these
clinical factors may then be correlated with specific magnetic resonance
and functional imaging changes. Conversely, non-melancholic disorders
may be characterised by certain cognitive attitudes, interpersonal be-
haviours and the absence of psychomotor change, and may be found to
present a differing pattern of clinical and neurobiological correlates.
Attempts should also be made to (i) identify possible hierarchical re-
lationships within such data sets and (ii) partial out the effects of key
co-variates such as gender and age of onset. For example, in melancholia
we have proposed that psychomotor disturbance is a specific addition to
other clinical phenomena (e.g., mood disturbance, psychotic features and
vegetative symptoms), while, in non-melancholic disorders, one might
assert that mood disturbance is sufficient to explain the cognitive and
interpersonal phenomena. That is, the extent to which any pattern of
associations can be explained by a smaller number of constructs needs
to be formally examined. Clinical practice has proposed a range of
"independent" constructs which may not behave separately at the neu-
robiological level. While it is unlikely, however, that there are multiple
independent constructs, it is equally unlikely that all behavioural con-
structs are present in all depressive disorders. Further, some constructs
may result more directly from underlying neurobiological processes and
lend themselves readily to current investigative techniques (e.g., psy-
chomotor change may be a direct consequence of structural changes in
the basal ganglia of older depressives). By contrast, other constructs,
such as cognitive attributes or interpersonal style, are more likely to
represent the outcome of complex interactions between temperament,
developmental experience and the demand characteristics of the current
social environment.
46 /. ffickie

Recording Behavioural Constructs


Data are likely to be useful in latent statistical models only if they closely
reflect the nature of the behaviour under examination. For example, at
a global level, psychomotor retardation may be said to occur in severe
depression, Parkinson's disease, subcortical dementia and negative forms
of schizophrenia. However, psychomotor retardation in depression may
have quite specific features, such as non-responsiveness to affectively
laden interpersonal or contextual cues. Such a feature may not char-
acterise the motoric abnormalities observed in Parkinson's disease or
subcortical dementia. The tendency to assume that global ratings of
specific constructs will suffice may, therefore, introduce unacceptable
experimental error. In this setting, it is imperative to develop appro-
priate instruments which characterise closely, and record reliably, the
phenomena of interest.
An emphasis on examining the relationships between specific con-
structs within homogeneous patient groups, rather than collecting broad
symptom data from heterogeneous clinical samples, would also permit
researchers to build more readily on other medical and basic neuro-
science research. For example, a focus on psychomotor change in pa-
tients with severe depression encourages a more direct examination of
developments in the relevant neurosciences detailing structural and func-
tional aspects of the basal ganglia, and their subcortical and cortical
projections (Alexander, DeLong and Strick 1986; Alexander, Crutcher
and DeLong 1990; Gerfen 1992; Joel and Weiner 1994).

Integrating Human and Animal Research


Animal models of depression have tended to focus on specific behaviours
that are assumed to be analogous to specific cognitive or behavioural as-
pects of depression in humans (e.g., learned helplessness). The capacity
to integrate information from animal models may, however, be advanced
by recording more precisely relevant behavioural and cognitive compo-
nents in humans. For example, while there is unlikely to be an ani-
mal analogue of the broad construct of major depression, psychomotor
change can be induced in animals by a variety of pathophysiological
processes (e.g., post-infective states [Hart 1988]). That is, while it may
not be possible to describe an adequate animal model for the full spec-
trum of depressive disorders, it may well be possible to develop a close
animal analogue of psychomotor change, and then conduct detailed in-
Behavioural Constructs in Melancholia Research 47
vestigation of the pathophysiological processes which may induce and/or
mediate that phenomenon.
Conversely, rapid developments in the neurosciences, particularly at
the molecular and cellular levels, will result in the description of a wide
variety of potential pathophysiological processes. Such new develop-
ments may not correlate with the clinical syndromes currently described
in neuropsychiatry. They may, however, be able to be more precisely
evaluated in terms of their contribution to simpler behavioural con-
structs. For example, if one wishes to study the relationship(s) between
neurochemical receptor subtypes and behaviour, one may employ a wide
range of specific behavioural observations (including simple factors such
as appetite, weight change and sleep-wake cycle disturbances as well
as more complex phenomena such as psychomotor and social activity)
rather than global concepts such as "depression."

Aetiological Classification Models


All clinically driven approaches to classification are made redundant
eventually by the identification of aetiological processes. Such aeti-
ological processes may be revealed by serendipity or by planned re-
search in neuropsychiatry. At this stage, encouraging neurobiological
correlates of a range of neuropsychiatric disorders are being identified
(e.g., apolipoprotein E4 (ApoE4) genotype as a risk factor to late-
onset Alzheimer's disease [Strittmatter et al. 1993; Saunders et al. 1993;
Corder et al. 1993; Poirier et al. 1993]), but the specificity of such associ-
ations needs to be closely evaluated. Planned neuropsychiatric research
needs to consider a wide range of possible risk factors for each disorder
and/or each behavioural construct. Certain risk factors are likely to pre-
dispose to a wide range of behavioural phenomena and may suggest that
current systems of sub-classification lack validity. Of most concern, an
overemphasis on detecting associations between diffuse diagnostic cat-
egories and potential biological markers may prematurely limit further
investigations of such markers. This destructive process is well demon-
strated by the historical demise of the DST as a specific marker for
melancholia. By contrast with that literature based on the diagnosis
of major depression, we have demonstrated a robust relationship be-
tween the DST and psychomotor change (see Chapter 8). When clear
pathophysiological approaches are eventually described, then the spe-
cific resultant behavioural patterns will also be revealed. It is predicted
that such resultant behavioural patterns will more closely approximate
48 /. Hickie
those described by the latent statistical techniques proposed here than
the clinical disorders formalised by current diagnostic systems.
It can be argued that any attempts to link paradigms or direct obser-
vations from clinical psychiatry or psychology with specific brain regions
or pathophysiological processes will not be fruitful, with real advances
occurring only after the functional connections at the neural and bio-
chemical levels between various brain areas have been established. Prom
this perspective, no real advances in classification will occur prior to key
developments in the basic neurosciences. An alternative view, however,
proposes that such advances themselves, at least where pertinent to the
neuropsychiatric disorders, will always require a sophisticated under-
standing of the demand characteristics of the surrounding environment.
That is, knowledge of brain processes per se will not permit predic-
tions of specific behaviours. For example, unravelling the cortical and
subcortical correlates of observed psychomotor change may depend on
understanding the way in which specific and non-specific environmental
stressors (e.g., death of an intimate, severe illness) interact with any
biological substrate to result in observed clinical features such as psy-
chomotor agitation, stereotyped verbal responses or social withdrawal.

Current M D U Research
Our research endeavours with melancholia can be viewed as having met
a number of these steps for planned neurobiological research utilising be-
havioural constructs. First, we have identified a range of key clinical con-
structs (e.g., psychomotor change, psychotic features) and, second, we
have developed a sensitive and reliable instrument for rating psychomo-
tor phenomena in patients with depressive disorders (CORE system).
Third, we have selected appropriate clinical samples for investigating
specific neurobiological phenomena (see Chapters 8 and 17 detailing the
DST and magnetic resonance imaging [MRI] studies). Fourth, we have
utilized latent statistical techniques to examine the patterns of associ-
ation between the behavioural and symptom data collected and, where
possible, compared such mathematical models with independent neuro-
biological and clinical data (see Chapter 6). As yet, we have not fully
utilized the potential of such techniques by employing relevant neuro-
logical control groups and/or extending the breadth of our clinical and
neurobiological data base. Further, we have not utilized measures of
psychomotor change in animals to clarify possible pathophysiological
processes. Having successfully developed an appropriate measure of a
Behavioural Constructs in Melancholia Research 49
key behavioural feature of severe depressive disorders, we now plan to
employ it imaginatively, in combination with other key measures, clinical
outcome variables and an enriched array of neurobiological correlates.

Neuroanatomical Models of Depression


It is relevant to compare our MDU approach to the study of melancholia
with other current models for severe depressive disorders. A number of
models linking data from neuroimaging (notably MRI), and neurochemi-
cal studies, with current concepts of the complex circuitry linking frontal
and striatal structures (Alexander et al. 1986; Alexander et al. 1990; Ger-
fen 1992; Joel and Weiner 1994) have been proposed (see Chapter 15).
Krishnan (1993b) has most clearly articulated such a model for late-life
depressive disorders. Given the close clinical similarities between late-
life depression and the traditional notion of melancholia (i.e., both hav-
ing observed psychomotor change, demonstrable cognitive impairment,
frequent psychotic features, high rates of medical co-morbidity and pref-
erential response to physical treatments), it is reasonable to assume that
this model also provides a working hypothesis for melancholia.
Krishnan's model is essentially a stress-diathesis hypothesis whereby
frontal white matter and basal ganglia lesions (occurring secondary to
vascular or genetic factors) provide the physiological substrate. White
matter lesions are viewed as producing damage to the glutamatergic pro-
jections from the cortex to the striatum, while lesions of the basal ganglia
result in damage to the 7-aminobutyric acid (GABA)ergic component
of the fronto-striatal systems. Other possible mechanisms include sec-
ondary damage to the noradrenalin system from subcortical or pontine
white matter lesions. Given the large reserve capacity of CNS mecha-
nisms, such effects do not become clinically manifest until the individual
is required to respond to some additional stressor. This model integrates
current knowledge regarding cortical and basal ganglia circuitry, MRI le-
sion findings and neurochemical perspectives. It places central relevance
not only on the role of the basal ganglia in determining sensorimotor as-
pects of movement, but also on their increasingly apparent role in the
traditional limbic functions of mood and cognition.
Cummings (1993) has similarly sought to integrate functional and
structural neuroimaging data to propose a "neuroanatomy of depres-
sion" which links specific (and restricted) brain regions with specific
clinical phenomena. Again, changes in the fronto-temporal (limbic) cor-
tex and caudate nucleus are proposed as critical features. Of relevance
50 /. Hickie

to melancholia, Cummings proposes that the motor manifestations (af-


fecting posture, gait and speech) of depression are similar to those in
Parkinson's disease, where such features appear to be mediated by the
basal ganglia (for posture, gait and speech) and the right frontal cortex
(for speech).
Mayberg (1993) has proposed a unifying model for depression in
patients with known basal ganglia disease which is also relevant to
melancholia researchers. On the basis of neuroimaging studies, she hy-
pothesises that selective disruption of cortico-striatal and baso-temporal
limbic pathways results in characteristic metabolic defects in the orbito-
frontal and temporal cortex and corresponding depressive symptomatol-
ogy. Possible sites for such disruptions include: cell loss from the frontal
cortex; degeneration of the basal amygdaloid nucleus; basal ganglia-
thalamic-cortical pathways degeneration secondary to caudate nucleus
degeneration; primary degeneration of the ventral tegmental area, with
disruption of dopamine projections to the mesolimbic frontal cortex;
and/or secondary changes in brain stem monoamines from dysfunction
of the orbito-frontal cortex.
In Carroll's (1991, 1994) three-component model of manic depression,
various key symptom complexes (anhedonia, dysphoria and psychomotor
function) are linked with proposed behavioural systems (reinforcement-
reward, central pain and psychomotor regulation) and potential neuro-
chemical substrates (noradrenergic, serotonergic, adrenergic and dopami-
nergic). Specifically with regard to psychomotor regulation, the key
anatomical substrate includes the frontal cortex, basal ganglia, cerebel-
lum and thalamus, with the subcortical striatal system having executive
control over motor programming. The neurotransmitter correlates of
this system are primarily dopamine (having an accelerating role) and
acetylcholine (having the reverse effect). This model is most consistent
with the general theme of this chapter - namely, that neurobiological re-
search would be greatly assisted by a focus on key behavioural constructs
rather than on formal diagnostic categories.

Research Questions in Melancholia


In light of these models and other relevant risk factor data, it is now
possible to generate more elaborate hypotheses linking neurobiological
factors with the clinical construct of melancholia. In doing so, the likely
influence of dopaminergic mechanisms in depressive disorders charac-
terised by mood and movement disorders, and the possible confounding
Behavioural Constructs in Melancholia Research 51
effects of age, psychotic features and bipolarity, need to be carefully
elaborated.

Dopaminergic Mechanisms
Dopaminergic mechanisms are of specific relevance to those diseases
(e.g., Parkinson's disease) characterised by movement disorders. Dopa-
mine is also relevant to mood state, as those antidepressants with signifi-
cant dopaminergic activity (e.g., nomifensine, bupropion) are efficacious,
and may have specific benefits for patients characterised by subjective
anergia and observable psychomotor slowing (Brown and Gershon 1993).
Chronic treatment with traditional tricyclic agents, monoamine oxidase
inhibitors and ECT may result in increased dopamine release, secondary
to decreased sensitivity of inhibitory pre-synaptic dopamine receptors
(Antleman and Chiodo 1981). Cerebrospinal fluid studies have associ-
ated reduction in the dopamine metabolite homovanillic acid (HVA) with
the retarded style of depression (van Praag, Korf and Schut 1973). As
proposed by Carroll (1991; 1994), it is therefore possible that dopamin-
ergic mechanisms (or the balance between dopamine and acetylcholine)
are responsible for both mood changes and the movement disorder in
patients with certain sub-types of depression.
Dopamine has also long been implicated in psychotic disorders, with
the efficacy of antipsychotic agents being generally positively correlated
with their capacity to bind D2 receptors, and dopaminergic agents being
identified as precipitators of psychotic features. Consequently, an appar-
ent paradox arises with regard to the relationships between dopaminergic
activity and melancholia, as psychotic features (implied dopaminergic
over-activity) and psychomotor change (implied dopaminergic under-
act ivity) often occur concurrently. A number of factors are relevant,
including:
(i) Carroll (1991; 1994) posits the notion that one neurotransmitter
abnormality (acting in one region or globally) is unlikely to account
for all behavioural features. Therefore, the relative state of other
key neurotransmitters in critical regions (e.g., acetylcholine in the
basal ganglia; serotonin and acetylcholine in the amygdala and as-
sociated cortical regions) will also determine the final combination
of clinical symptoms presented;
(ii) The relationship between D2 receptor blockade and antipsychotic
activity has been challenged by functional imaging and clinical stud-
52 /. Hickie

ies (Pilowsky et al. 1992) which demonstrate that atypical antipsy-


chotics (notably clozapine) are effective antipsychotics, even though
they have low affinity for D 2 receptors. Consequently, other neu-
rochemical substrates for psychosis require careful examination, es-
pecially as clozapine has demonstrated action at various serotonin
(5-HT2, 5-HT3) and other dopamine (D4) receptors; and
(iii) Differential dopamine activity at various neuroanatomical sites may
be relevant - with under-activity within nigrostriatal pathways me-
diating largely motor changes, while over-activity within other meso-
limbic and/or mesocortical pathways may account for concurrent
psychotic features.

Age and Melancholia


Although most classical melancholic disorders are encountered in older
patients, the specific role of ageing awaits clarification. Age may sim-
ply have a pathoplastic effect, with subjective anergia and cognitive
complaints being more characteristic of younger patients, and objec-
tive psychomotor slowing and demonstrable cognitive decline being more
characteristic of older patients. Further, interactive effects between psy-
chosocial and neurobiological risk factors may be evident. For example,
psychomotor agitation (rather than retardation) has been linked histor-
ically with specific pre-morbid (anxious, obsessional) personality traits,
suggesting that when such persons become depressed in later life they
preferentially develop a particular type of psychomotor change (Lewis
1967). A research system which allows differentiation of the various
types of psychomotor change and concurrent assessment of such pre-
morbid clinical traits permits such speculative hypotheses to be evalu-
ated.
The most likely effect of ageing is that it constitutes a genuine risk fac-
tor to melancholia via two paths. First, the cumulative loss of neurons
common to all persons with age (and particularly as a result of effects on
frontal structures) may result eventually in disruptions of critical subcor-
tical and basal ganglia circuits and create a key neuroanatomic vulner-
ability (Krishnan 1993b). Secondly, this normal ageing process may be
accelerated in certain individuals by specific genetic factors (e.g., pres-
ence of the Apo E4 genotype), prior regional neuroanatomical insults
(e.g., fronto-temporal head injury), relevant CNS disorders (e.g., multi-
ple sclerosis), or systemic disorders which affect the CNS, with particular
Behavioural Constructs in Melancholia Research 53
emphasis on those disorders affecting the vasculature (e.g., hyperten-
sion, hypercholesterolaemia, elevated homocysteine, anti-phospholipid
antibody syndrome). The disorder may then become clinically evident
either in the face of individually relevant life stressors or when the neu-
ronal loss reaches such a critical stage that no specific or major precip-
itants are required. If this framework is valid, an increased prevalence
of melancholia in older patients, and a markedly increased rate and ear-
lier age of onset in those persons with specific risk factors would be
expected. Further, one would predict that those with early-onset non-
melancholic depressive disorders (presumably predisposed to by other
genetic, temperamental and/or other psychosocial factors) would not go
on to develop melancholic syndromes in later life.
Classic melancholic syndromes do, however, develop in some younger
patients. In our clinical sample, such patients can present the full range
of psychomotor and cognitive impairments. Typically, such younger pa-
tients have a history of bipolar disorder and frequently report periods
of anergia and social withdrawal prior to their first treated affective
episode. Such clinical observations suggest that the classic melancholic
features (psychomotor change, cognitive impairment, psychosis) are not
restricted to older persons. MRI studies of young bipolar patients need
to be viewed cautiously, but it does appear as if areas of subcorti-
cal hyperintensity are over-represented, suggesting clinically significant
neuroanatomical changes (Dupont et al. 1990). Whether a significant
proportion of these younger patients have experienced other neurobio-
logical risk factors which have resulted in regional neuronal loss compara-
ble with that due to ageing requires close cross-sectional and longitudinal
evaluation.
Importantly, a significant proportion of patients with early-onset bipo-
lar disorders have strong family histories of manic depressive illness but
do not appear to have relevant structural risk factors. We have demon-
strated that the depressions presented by bipolar patients are clinically
indistinguishable from those with unipolar melancholia (see Chapter 11).
Although the relevance of structural risk factors here requires system-
atic longitudinal imaging and clinical investigation, it does suggest that
an alternate hypothesis is required for the aetiology of melancholia in
young bipolar patients. Clearly, familial transmission factors suggest a
genetic predisposition model, where the liability is functionally rather
than structurally mediated, though the anatomical localization of such
dysfunction may correspond with the regions identified in MRI studies
of older patients (notably basal ganglia, subcortical white matter and
54 I. Hickie

prefrontal cortex). At the most simplistic level, a model of inherited


dysfunctional dopaminergic and/or cholinergic systems could account
for the triad of psychomotor change, mood and cognitive impairment.

Psychotic Features
The possible relationships between melancholia and psychotic depression
will be explored in detail in Chapter 12. Importantly, psychotic depres-
sion is associated with the most severe forms of psychomotor disturbance
and, compared with non-psychotic melancholia, appears to be more
specifically associated with profound non-interactiveness and severe ag-
itation. Given other important clinical differences (e.g., preferential
response to ECT), it is likely that psychotic depression is associated
with at least some unique neurobiological features (e.g., dopaminergic
over-activity within mesocortical pathways). Our preliminary MRI study
(Chapter 17) suggests that relationships between psychomotor change
(CORE scores) and subcortical white matter changes are different in
patients with and without psychotic features. At this stage we can-
not determine whether psychotic depression simply represents the most
severe form of melancholia or a distinct sub-class (with its own dis-
tinct clinical features due to unique pathophysiological processes). Our
own view is that psychotic depression is a distinct sub-class of melan-
cholia, but careful clinical, longitudinal and neurobiological study of
large cohorts is required to determine the underlying relationships be-
tween psychotic depression, melancholia and non-melancholic depressive
disorders.

Bipolarity
Although our own studies have tended to suggest that patients in the
depressed phase of bipolar disordes are phenotypically very similar to
patients with unipolar melancholia (see Chapter 11), other clinical (e.g.,
earlier age of onset, prolonged sleep, profoundly reduced motor activity
in young patients); neurobiological (e.g., differential circadian rhythm
disturbance, with bipolar patients demonstrating phase advancement
compared with phase delay in unipolar patients); and treatment (e.g.,
specific roles for lithium and anti-convulsants) data have tended to sug-
gest that bipolar patients may need to be treated as a separate sub-class.
Consequently, the careful recording of the likelihood of bipolarity needs
to be included in studies detailing melancholic disorders.
Behavioural Constructs in Melancholia Research 55
Conclusion
In summary, working hypotheses for the syndrome of melancholia em-
phasise the likely role of (inherited or acquired) neurochemical deficits
within localized fronto-striatal circuits (Krishnan 1993a; Chapter 15;
Cummings 1993). Using such hypotheses, specific cross-sectional and
longitudinal research questions for melancholia can now be stated. Key
questions include:
(i) What are the specific relationships between structural (MRI) and
functional (including "at rest," cognitive activation and dopamine
receptor studies) imaging studies in patients with melancholia?
(ii) Are such inter-relationships affected by other confounding factors
including current age, age at onset of affective disorder, hospital-
ization, presence of psychotic features and the presence of a bipolar
disorder?
(iii) Is the distinction between early- and late-onset major depressive
disorders suggested by earlier MRI and risk factor studies appli-
cable to melancholia (and psychotic depression)?
(iv) Do a large proportion of those with late-onset melancholia, with
associated MRI changes, eventually develop irreversible cognitive
change?
(v) Are MRI changes in those with early-onset melancholia indicative
of certain predisposing genetic or environmental factors?
(vi) What range of depressive disorders do patients with neurologi-
cal disorders affecting key fronto-striatal circuits actually develop,
and, if they do develop the classical behavioural features of melan-
cholia, are these then associated with comparable MRI and func-
tional imaging deficits?
(vii) What specific relationships exist between the melancholic con-
structs (mood change, psychomotor change, cognitive impairment
and psychotic features) and dopamine and/or acetylcholine recep-
tor function in nigrostriatal and mesolimbic/mesocortical struc-
tures?
(viii) Do those with early-onset melancholia go on to develop MRI
changes over time and, if so, are such changes correlates of other
known risk factors, of chronic depression or of some treatment fac-
tor (e.g., chronic administration of antidepressants, antipsychotics
or ECT)?
(ix) Do specific vascular and/or genetic risk factors predispose to all of
the key melancholic constructs, or do they predispose specifically
56 /. Hickie

to psychomotor change, cognitive impairment, psychotic features


and/or a bipolar course?
(x) Is the presence of psychotic features in patients with melancholia
indicative of a differential pattern of psychosocial and/or neurobi-
ological risk factors and, thus, generally supportive of the notion
of a unique sub-class?
(xi) Is the presence of a late-onset bipolar longitudinal history indica-
tive of specific neurobiological risk factors, and do such differences
substantiate separation from bipolar patients with early-onset dis-
orders?
The capacity to answer such research questions depends on the inves-
tigator's capacity to utilize data concurrently from behavioural, clinical,
neurobiological and longitudinal sources. Although intrinsically more
complex (conceptually, logistically and analytically), utilisation of data
from a series of behavioural and other clinical domains, in combina-
tion with detailed neurobiological findings, may well provide more pro-
found insights into the aetiology of complex psychiatric disorders such as
melancholia and psychotic depression than more traditional diagnosis-
based approaches.
4
Issues in Classification:
IV. Some Statistical Aspects
DUSAN HADZI-PAVLOVIC

In the 1960s the unitary/binary debate about the classification of de-


pression was complemented by a debate about the capacity of statistical
methodologies to help in deciding nosological questions.
A common nosological pursuit is to examine whether a disorder is
categorical or dimensional. In the broad domain of biology the task of
deciding mathematically between categories and dimensions goes back
to the geneticists of the late nineteenth century and Pearson's (1894) in-
troduction of the method of moments for deriving a solution for what we
now call the finite mixture model. Two persisting central concerns have
been the capacities of statistical techniques to distinguish between these
possibilities, and how the data we collect can mislead interpretation.
While the generic arguments were most vigorous initially in the con-
text of depression, they touch all psychiatric disorders, and are currently
perhaps most active in the area of personality research (e.g., Gangestad
and Snyder 1985; Grove and Andreasen 1989; Meehl 1992).

The Nosological Question in Depression. This can be construed


as a set of arguments about
• whether depression lies on a continuum with normality or is a sep-
arate entity;
• if depression is not on a continuum, whether the class of depressives
is a continuum or comprises a number of discrete disorders; and
• if depression is a discrete entity, whether the distinguishable disor-
ders are best described using categorical or continuous description.
As noted by Eysenck (1970) "[T]here are two, not one, problems in-
volved, relating (a) to the unitary or binary nature of depression, and
(b) to the categorical or dimensional nature of these illnesses."

57
58 D. Hadzi-Pavlovic

The unitary view posits that depression is a single disorder and, as


it worsens, the number of symptoms increases and certain symptoms
become more likely to appear. So the presence of delusions, for example,
is viewed as indicative not of a different disorder but of severity having
reached the point at which delusions start to manifest. The binary view
does not exclude a dimension of severity, but it does assert that certain
manifestations of the depression (e.g., delusions) arise from a different
disorder (i.e., from different causes at least at some points on the causal
chain).
The phrase different disorder might not be appropriate if what we are
talking about is more a sub-type. While the distinction between a sub-
type and a different type is probably important, the mechanics of making
that distinction, and the more immediate consequences of so doing, are
not at all obvious, especially if the "types" share features. For example,
if the view that there are sub-types of Parkinson's disease is correct -
where those sub-types are thought to be the sequelae of differences in
particular structures - it is still likely that we would want to retain
the higher-level disorder - Parkinson's disease. Similarly, in Chapter 7
we provide evidence suggesting that psychotic depression shares a num-
ber of features with non-psychotic melancholia, but that at some point
additional features intrude to produce the form (the sub-type) of melan-
cholia peculiar to psychotic depression. It probably does not require a
lot of data to shift one's view from a "different disorder" interpretation
of this condition to a "sub-type" interpretation. One reason that such
distinctions are potentially so labile is that statistical methodology re-
ally provides only weak support for favouring one model (one taxonomy)
over another. Bartholomew (1987) warns "that a latent variable with a
continuous normal distribution and one in which all the prior probability
is concentrated at K points [can] lead to very similar data matrices."
Another source of difficulty is that the processes (or the aspects of
a process that we can measure) underlying a disorder can be either
discrete or continuous. If depression results from an increased rate of pre-
synaptic neurotransmitter re-uptake, then any demonstration of uptake
being continuous does not reject the possibility of distinct disorders.
Returning to the analogy of Parkinson's disease, we do not reject it as
a discrete entity just because we can measure parkinsonian rigidity on
a continuous scale that starts with "no rigidity present," or because we
could, in theory, plot the continuous decline in dopamine-related neural
structures that accompanies the emergence of the discrete disease.
Some Statistical Aspects 59
If neurotransmitter re-uptake is an epiphenomenon or side effect, if
rigidity is an effect well away from the central disturbance, then the
nature of these might well not count as evidence at all in any pursuit
of dimensional versus categorical disorder status. Deciding what is the
appropriate domain to measure and the appropriate procedures for nu-
merical taxonomy are therefore not trivial issues. Some illustrations of
the complexities will now be noted.

Testing for Categories or Sub-types


The Argument from Factor Analysis
Despite alternative interpretations by many writers (see Chapter 2),
the primary roles that factor analysis (FA) and principal components
analysis (PCA) have taken is in the argument that certain patterns of
the factor loadings are held to be consistent with the binary or unitary
view.
The most extended account of how the unitary and binary views could
be distinguished was given by Eysenck (1970) in terms of FA. Assuming
that the measures being factored are appropriate (or, as put by Eysenck
[1970], "items thought on clinical grounds to be relevant to both putative
types of depression"), then,

(i) under both views, the first factor would be a general factor;
(ii) if the Unitarian view is correct, then this first factor should be suf-
ficient to explain the inter-correlations; or
(iii) if the binary view is correct, then at least one more factor would be
required. This second, bipolar, factor would be at least as important
as the first, and would have one pole defined by endogenous features
and the other by non-endogenous ("reactive") features.

Maxwell (1971) criticised Eysenck's account (implicitly) and a similar


argument by Hamilton (explicitly) when he stated, "It is a property of
factor analysis ... that the signs of the loadings on the second factor will
be roughly half positive and half negative. In other words the second
factor will divide the variables into two contrasting groups." This will
usually, but not necessarily, be true of unrotated, orthogonal factors of
a correlation matrix where the correlations are positive. One way to
conceptualize the issue is to view a PCA as suggested by McDonald
(1985). Usually we think of the first component as that combination
60 D. Hadzi-Pavlovic
which accounts for the most variance. McDonald, however, suggested
viewing the first component as that combination which most looks like
each of the individual variables. If we have four variables, say, A, J3, C
and D, then intuitively, the combination of these which, simultaneously,
looks most like A, and most like B and so on, will most likely be the
average or sum of the variables (i.e., A+B+C+D). The first component
usually turns out to be such a sum. The second component, being
uncorrelated with the first, needs to look like those parts of the variables
which do not resemble the first. Again it seems intuitive that, having
found as much of the similarity that is expressible as sums of variables,
we would now need to turn to differences between variables. The second
component typically comprises positively and negatively weighted items
(differences) - what is usually referred to as a bipolar factor.
The critical thing then has to be "the items at the poles and not the
emergence of a bipolar factor."

The Argument from Bimodality


Potentially, one of the most persuasive arguments for demonstrating cat-
egories (if present) is the demonstration of bimodality in the frequency
distribution associated with the generic disorder. As noted in Chap-
ter 2, some have placed a great weight on bimodality. Kendell (1968)
wrote that, in deciding whether "endogenous and neurotic depressions
are separate entities . . . [or] . . . opposite ends of a continuum. . . . Only
the demonstration of a bimodal distribution curve for some property of
an unselected sample of the whole population is adequate to distinguish
the two possibilities." Others, however, have felt that there are too
many problems with demonstrating bimodality to make it a criterion
worth pursuing (e.g., Grove and Andreasen 1989).
Bimodality should not be confused with "looks bimodal." In trying
to decide whether some measure of depression has a bell-shaped (or an
Elle-shapedf) curve we ought not to be seduced by appearances, since
unimodal shapes can appear even when we have a genuine mixture from
two populations. There are many factors which can produce spurious
unimodality and bimodality alike (e.g., Murphy 1964; Grove and Telle-
gen 1985). Bimodality, then, is better thought of as the successful fitting
of certain statistical models, and it is not appropriate to focus on visual
proof.

f This is a parochial Australian reference to a celebrated beauty of the day.


Some Statistical Aspects 61
A good example of examining for bimodality is the testing of a single-
gene explanation of blood pressure (this is based on Roeder 1994). With
two alleles, under a simple dominance model there are three distinct
genotypes which give rise to two phenotypes, Q\ and #2- The probabili-
ties of the two alleles determine the probabilities of the two phenotypes.
Individuals of type d\ will have a mean blood pressure of M\ and the
others M2. Observed blood pressure, however, will show variation -
due to measurement error and other causes of blood pressure - so that
individuals will fall above or below their expected mean. If we assume
that error and the other causes are distributed normally, then we would
expect the 6\ (02) phenotype individuals to have levels distributed nor-
mally, with a mean of M\ (M2) and variance a\ (c^). The totality of
observed blood pressure, therefore, will be a mixture of samples from two
normal distributions, each of which arises from a distinct aetiology, and
with the proportion in each distribution determined by the frequencies
of the alleles. This example illustrates
(i) how distinctly categorical conditions (different aetiologies) can lead
to a continuous phenomenon (blood pressure) and how they can be
separated out; and
(ii) how the whole argument is strengthened by being able to posit a
clear aetiological mechanism.
In principle, there is no reason why we cannot replace blood pressure
with neuroticism or depression and demonstrate sub-populations. In
psychiatric research this is likely to be the start of an iterative process:
demonstrate bimodality; use the categories defined by the analysis to
guide the refinement of measures and aetiological hypotheses; and re-
test the bimodality hypothesis.

Mixture Models. For the example given above there are a number of
methods for estimating the means and variances of, as well as the pro-
portion coming from, each distribution (McLachlan and Basford 1988).
There also exist statistics which allow us to test whether the addition
of another distribution results in a significantly better fit to the data.
The initial null hypothesis is that the data are consistent with a sample
from a single distribution. If this hypothesis is rejected, that is, one
distribution in not sufficient, then one can test if two distributions are
sufficient and so on. Thus Roeder (1994) was able, using her methodol-
ogy, to reject the simple dominance model, showing the data to be more
consistent with an additive model (i.e., one with three phenotypes).
62 D. Hadzi-Pavlovic

Discriminant Scores. The classic attempt to show bimodality was


KendelPs (1968) examination of scores from a discriminant function
analysis (DFA). A DFA differs from the other methods mentioned here,
in that group membership (diagnosis) is already known (more precisely,
this knowledge is used in the analysis), with the DFA scores representing
that linear combination of variables for which these groups (in this case
endogenous and neurotic depressives) are maximally separate.

Factor Scores. Factor scores represent scores on the unmeasured fac-


tors or latent variables that are uncovered by the FA. In principle, we are
just as entitled to expect bimodality in a latent variable as in a measured
variable, but we do need to ensure that bimodality is possible, on a case
by case basis, for each different kind of latent variable model. Maxwell
(1971) claimed that FA ought not to produce bimodality because fac-
tor scores are "weighted sums [which] will tend strongly to normality
... in view of the central limit theorem." It is not that straightforward,
since DFA scores are also weighted sums and Maxwell accepts that these
scores can result in bimodal distributions (and indeed often will). Sub-
sequently, he outlines those circumstances under which bimodality can
appear in FA. These include comparability of scale for the items and
using the known diagnostic group memberships (as in DFA) to form
correlation matrices. While the conditions are strict they are not im-
possible to meet, and Maxwell's disposal of bimodality in factor scores
needs to be seen as model specific.

Multidimensional Mixture Models. When there are two or more


measures it is possible to model them as mixtures of multivariate nor-
mal distributions. Using two measures, bimodality is replaced by three-
dimensional "twin peaks." Such a model was used by Davis et al. (1988)
with biological markers of depression, to successfully identify the de-
pressed and control groups.

Latent Class Models. The examples so far have involved purported


classes of individuals measured on a continuous variable. If the measure
assesses data categorically (e.g., "no, yes", "nil, mild, moderate, severe")
then it is possible to define what are usually called latent class models,
though the term "latent class" is somewhat misleading, since a mixture
of normal distributions also involves latent classes. These models can
be thought of as mixture models, so, for example, if the data are all
binary, we have a multivariate mixture of binomial variables. The output
Some Statistical Aspects 63
of a latent class analysis (LCA) consists of (i) the proportion of the
sample which comes from each class; and (ii) for all the variables, the
probabilities of each possible response, given that an individual comes
from a specific class.
A number of mixture-like methods have been developed whose prime
focus is the taxonomic problem of identifying disorders (see Golden
1991). One of these (Golden 1982) was used in our original CORE
analyses (Parker et al. 1990).

Misleading Samples and Misleading Data


The possibility that the data which we collect might - even if appropri-
ately analysed - result in bimodality which is artefactual or unrepresen-
tative must also be conceded.

The Fallacy of the Unmeasured Middle. As mentioned in Chap-


ter 1, both Mapother and Ross recognised the need to survey the full
spectrum of depression, not merely the pure ("white" and "black")
types, but in particular to examine closely the greys in the middle and
see whether they were a Unitarian bridge or confusion clouding a bi-
narian gap. As Kendell (1968) put it, "Unless the starting material is
unselected, bimodal distribution curves, for whatever parameters, prove
nothing." Exactly what should constitute the starting material is some-
thing that has to be thought through. Kendell (1968) would appear to
be content with a patient population, whereas Grayson (1987) sees a
community sample as more appropriate.

The Fallacy of the Over-Measured Middle. A key argument put by


Grayson (1987) is that an inappropriate choice of items to form a rating
scale can make a set of individuals who truly lie along a continuum to
appear to be from distinct entities. For each item there is a point along
the severity dimension — the threshold — at which an individual is
more likely to endorse the item than not endorse it. Grayson shows how
a set of items whose thresholds are all around the middle of the severity
dimension will result in people below the middle "all" answering "no"
and those above answering "yes," thus producing what will appear to
be two classes of individuals.
64 D. Hadzi-Pavlovic
Conclusion
In this book we pursue the hypothesis that melancholic depression can be
distinguished, on the basis of clinical features only, from a residual non-
melancholic class. The objective of this chapter has been to indicate that
any attempt to demonstrate or reject the existence of separate classes
is open to multiple errors emerging principally from measurement and
analytic limitations, though subsequent interpretation or misinterpreta-
tion can add to the potential for error. In our empirical research we have
tried to respect the measurement issues and to use statistical methods of
a more appropriate kind - primarily models which assume latent classes
and allow either continuous or categorical variables to be analysed. We
have also found it fruitful to take the statistical solutions quite seriously
and pursue, through the data or through case notes, their surprising or
even counter-intuitive "suggestions." The final stage of interpretation
remains. While we certainly favour a modified binary view - melancho-
lia distinct from a residual heterogeneous class of depressive disorders -
the reader should consider all these issues when interpreting our data.
Part Two
Development and Validation of a Measure
of Psychomotor Retardation as a Marker
of Melancholia
Psychomotor Change as a Feature
of Depressive Disorders: Historical Overview
and Current Assessment Strategies
GORDON PARKER
HEATHER BROTCHIE

Irrationality and overt behavioural disturbance have been since Greek times
the two central features of madness. (Berrios 1985)
All writers agree as to the importance of it, but few define it or its manifes-
tations, leaving such phrases as "psychomotor retardation" or "inhibition" to
speak for themselves. (Lewis 1934)

Historical Overview
In this book we argue that psychomotor disturbance (PMD) is both
a key defining characteristic of melancholic depression and a reflec-
tion of its underlying pathophysiology. As noted by Rush and Weis-
senburger (1994), the terms "endogenous" and "melancholic" have had
differing meanings, including particular patterns of clinical features. In
their overview of nine "diagnostic operationalizations" of the melan-
cholic concept, psychomotor retardation was the only feature to be in-
cluded in all nine systems, while psychomotor agitation was included
in six. Nevertheless, psychomotor disturbance is principally viewed as
either (a) merely one of several features weighted to such a diagnosis;
or (6) more related to depression severity than to depression sub-typing
and, as such, to be secondary to the primary mood disturbance. We will
argue that PMD's utility as a specific defining feature of melancholia is
dependent on its measurement (requiring some precision in defining its
underlying domains and appropriate descriptors; and using behavioural
ratings rather than relying on symptom self-report data), and that such
measurement difficulties have resulted in lack of appreciation of its possi-
ble "necessary and sufficient" status as a melancholic sub-typing feature.
Before presenting research data to support those propositions, it is
appropriate to examine historical views about the relevance of PMD as

67
68 G. Parker and H. Brotchie
a clinical descriptor of melancholia. There is an immediate difficulty
in that over time the term "melancholia" has had varying meanings.
Jackson (1986) states that "melancholia" was the Latin transliteration
of a word used in ancient Greece to denote a mental condition involv-
ing prolonged fear and depression. It was derived in turn from another
word which translated into Latin as atra bills and into English as black
bile, with black bile being thought to determine the melancholic tem-
perament. Subsequently, the term "melancholia" was used over the cen-
turies to describe both a symptom state (when it might be applied to
any mood state of sorrow, dejection or despair) as well as a style of
temperament. During the seventeenth century the term "melancholia"
began to be restricted more to a depressive disorder than to a charac-
ter correlate, before losing prominence early in the twentieth century
(except in the case of involutional melancholia, which also subsequently
declined in use). The term reappeared as a sub-type of the major de-
pressive categories in the American DSM-III and DSM-III-R classificatory
systems. Here, its use both implied a severer form of depression and
a condition with clinical features corresponding to those held to define
"endogenous depression."
Such variable use of the term means that any attempt to attribute
awareness about or ignorance of the relevance of PMD to melancholia
over time becomes quite problematic. Thus, only a brief overview will
be provided, initially drawing extensively on one secondary source for
much of the historical material (Jackson 1986), with the intrigued reader
being referred to that extensive review.
According to Jackson, the Hippocratic writers in the fifth and fourth
centuries B.C. suggested a clinical picture of prolonged fear or depression,
of aversion to food, despondency, sleeplessness, irritability and restless-
ness as cardinal features of melancholia. Jackson suggested that the
earliest (and most persistent) descriptors addressed emotions, passions
or perturbations of the soul, and therefore ensured a focus on mood
state and temperament. By contrast, Berrios (1988) stated that in clas-
sical antiquity, "melancholia was defined in terms of overt behavioural
features ... [referring to] states of reduced behavioural output ... [and]
that symptoms reflecting pathological affect... were not part of the con-
cept." Thus, there are two contrasting views about the focus of early
descriptions. Certainly, extended descriptions of PMD are not evident
in Jackson's review of that period, where, at best, agitation may be
inferred by descriptors such as "mental anguish and distress" and re-
tardation by "silence" and related terms (Berrios 1988). Jackson states
Psychomotor Change: Historical Overview 69
that late in the second century A.D., Galen provided a broader clinical
description of melancholia, shaping definition of the clinical syndrome
for the next 1500 years. In essence, melancholia was then portrayed as
a chronic, non-febrile form of madness, with sufferers holding a single
delusional idea, and usually being fearful, sad, misanthropic and tired
of life. Symptoms of gastrointestinal distress were emphasised (as in
the earliest descriptions), particularly costiveness (constipation). Jack-
son noted that a second meaning for costive is "slow in action or in
expressing ideas, opinions," being derived from the Latin constipatus.
Jackson detailed how, in the second half of the seventeenth century,
some symptoms which had been associated with one of the three sub-
types of melancholia (flatuous, windy, hypochondriacal) became sepa-
rated out as hypochodriasis. While "basic clinical descriptions of melan-
cholia remained fairly stable" (Klein and Davis 1969), waning of the
humoral theory resulted in melancholia being interpreted in mechanistic
terms. Writers invoked the prevailing Zeitgeist as they sought to proceed
beyond descriptions of the emotional state in explaining evident PMD.
Thus, Jackson noted

(i) Pitcairn's explanation of a thicker than usual condition of the blood,


its excessive accumulation in the brain, and a sluggish circulation;
(ii) Hoffman's mechanical explanation (of "body fluid turning acidic
and fixed, its motions ... sluggish, and the person ... slow, timid
and sad";
(iii) Mead's neurocentric explanation involving an "aetherial nerve fluid,"
perhaps electrical, in which irregular motions could lead to melan-
cholia; and
(iv) Cullen's proposal that there was too little nervous fluid in melan-
cholia, with that depleted state explaining the observable anergic
behaviour.

Nineteenth-century accounts of melancholia reflected the emergence of


a strong descriptive and phenomenological orientation. Berrios (1988)
stated that symptom descriptions were made "possible by the avail-
ability of new psychological theories," with subjective experience being
used to define behaviour. While descriptions reflected the changed focus
in defining insanity by affectivity (Berrios 1985), reference to PMD ap-
pears richer than in earlier accounts. Jackson collected a number of such
descriptions to argue that the dejected state with its distressed preoccu-
pation and its concomitant "slowing" of mental and physical functions
70 G. Parker and H. Brotchie
was increasingly viewed as the descriptive core. Three examples will be
noted. First, Esquirol offered the following description: "The physiog-
nomy is fixed and changeless; but the muscles of the face are in a state of
convulsive tension." Secondly, Heinroth observed that patients "could
be slowed down, preoccupied, and deeply depressed," and evidencing
either a subdued silence or "sighing and lamentation," suggesting ex-
pressions of "retarded" and "agitated" depression. Thirdly, Maudsley
offered another behavioural description:
The two opposite spectacles of melancholy ... in the one case, he is so com-
pletely paralysed by it as not to make the least effort to help himself; in the
other, he instinctively puts his self-conservative energies into agitated struggles
of defence or escape.

The agitated form included "restless pacing up and down the room,
ceaseless lamentations and ejaculations of distress . . . perpetual moan-
ings and groanings . . . wringing of hands, beating of head and face or
rubbing of skin into sores" and with "agitation being its chief note, has
sometimes received the name of melancholia agitata" (Maudsley 1895).
By contrast, in describing "melancholia with stupor" or retardation,
Maudsley wrote that the patient "stands, sits, kneels, crouches or lies
all day in one almost motionless posture" and with "the mental tor-
por go other symptoms of sluggish vitality." Jackson (1986) quoted a
number of other observers. For instance, Krafft-Ebing wrote that melan-
cholia involved "a notable inhibition of the psychomotor aspect of the
mind," while Mercier documented "associated behavioural details," in-
cluding "the traditional symptoms of constipation and 'slowed down'
behaviour," as well as the condition of "active melancholia" - which
involved "exaggerated gestures, loud cryings and moanings," with the
latter suggestive of agitated depression.
By the eighth edition of Manic Depressive Insanity and Paranoia,
Kraepelin (1921) had brought mania and melancholia together as manic
depressive insanity. He termed the first group in his category of depres-
sive states "simple retardation," or "melancholia simplex," defined by
a sort of mental sluggishness; thought becomes difficult; the patients find
difficulty in coming to a decision and in expressing themselves. ... The process
of association of ideas is remarkably retarded ... they have nothing to say;
there is a dearth of ideas and a poverty of thought. ... They appear dull and
sluggish.

Kraepelin (1921) suggested that melancholia simplex was characterised


by the appearance of "a simple psychic inhibition without hallucinations
Psychomotor Change: Historical Overview 71
and without marked delusions," emphasising the paralysis of thoughts,
memory difficulties, and the sense of weariness and enervation. A sec-
ond group was characterised by delusions and hallucinations in addition
to "psychomotor retardation." Such patients with stupor, or "marked
psychic inhibition," might "lie mute in bed ... sit helpless before their
food ... hold fast what is pressed into their hand ... [and be] unable to
care for their bodily needs as should be." Additionally, Kraepelin listed
"agitated melancholia," whereby patients might show "anxious restless-
ness ... [or] beg for forgiveness, entreat for mercy, kneel, pray, pluck at
their clothes, arrange their hair, rub their hands restlessly, give utter-
ance to inarticulate cries" or "wander about, bewail and lament, often
in rhythmical cadence."
As noted by Jackson, the British psychiatrists Henderson and Gille-
spie shortly afterwards defined three depressive states (acute depression,
simple retardation and depressive stupor) building on Kraepelin's three
basic symptoms - "depression, difficulty in thinking, and psychomotor
retardation" (Jackson 1986).
After stating that two metaphors "stand out in the long history of
melancholia: being in a state of darkness and being weighed down,"
Jackson added a third, of "being slowed down," observing that its "roots
appear to be no older than the seventeenth century adaptations of me-
chanical theory, supported by the readily observed diminished and in-
hibited physical activity of so many melancholic patients." Jackson
then noted that the descriptor "slowed down" was used less, being rela-
tively uncommon in nineteenth-century accounts, only to return in the
twentieth century, being made explicit in phrases such as "slowness of
thought," "mental sluggishness" and "retardation of the mental stream,"
all contained in terms like "psychomotor retardation."
Thus, while in the late nineteenth century and the beginning of the
twentieth century the emphasis remained on the mood state as the es-
sential element in the clinical picture, retardation of thought and of
physical activity was increasingly viewed during that period as a key
aspect of the descriptive core (with agitation as a variant) but, because
PMD was conceived of as secondary to the mood state, there was the
potential for such a "behavioural sign" descriptor to be viewed as redun-
dant or minimised when clinical descriptions focussed on intra-psychic
phenomena and "symptoms." This potential was realised shortly after
and persists (we suggest) in contemporary systems with their focus on
symptoms and with minimal attention to signs.
Subsequent to Kraepelin's period of direct influence, PMD (in depres-
72 G. Parker and H. Brotchie

sion) was minimally addressed in a number of major phenomenological


texts (e.g., Jaspers 1913). At the cusp of Kraepelin's influence, Freud
published his influential paper on melancholia, with subsequent analysts,
theorists and, in particular, cognitive theorists focussing on cognitive
and intra-psychic features, thus further weighting a symptom-focussed
subjective approach. Thus, in developing the "objective-descriptive
school," and focussing on what might be observed, Kraepelin predictably
noted the behavioural manifestations of PMD. Subsequent theorists, by
focussing more on self-esteem, cognitive processes, dynamic formula-
tions and psychosocial aetiological issues, appeared less interested in
behavioural observations.
The decline in attention to behavioural manifestations of depression
is notable in English (as opposed to Western European) publications,
although that decline was redressed in part by Aubrey Lewis, perhaps
because of his awareness of the non-English European literature.
In the late twenties, Lewis studied 61 depressive patients, and his re-
port (1934) has been described by Kendell (1968) as "the most authori-
tative and comprehensive study of melancholia in the English language."
Lewis (1934) observed that, as opposed to self-description, another and
complementary method of judging mood was "by observing the patient's
expression and behaviour," and that the facial expression of emotion is
more easily recognized than described. He considered that analysis of it,
feature by feature (e.g., omega figures, Veruguth folds, and such details)
was less informative than a single glance or a word of description. He
suggested that behaviour, so far as depression goes, was a matter of pos-
ture and motor activity. For the most part, the posture was said to be
drooping and slack, unless there was also agitation, while motor activity
changes were evidenced by "retardation," "agitation" and "stupor."
Lewis (1934) described a number of reasons why retardation was dif-
ficult to rate from observation (e.g., "The conditions of hospital life are
artificial; the patient lies in bed during the worst part of his illness . . . " ) ,
and so made estimates from reports of patients and relatives. All but
two affirmed "fatigue or anergia," and Lewis observed that, in judging
retardation, "The incapacity is greater in the patient's report than it
appears in his behaviour." Thus, when estimates were based on "defi-
nite evidence in the patient's behaviour of a slowing of his activities,"
retardation was recorded as present in 19 of the patients, with emphasis
"being laid on slowness of movement and occasionally of talk, or, on
evidence of inhibition, as recognized in speech broken off before comple-
tion."
Psychomotor Change: Historical Overview 73
In relation to thought processes, as opposed to activity, Lewis made
reference to Kraepelin, who wrote:
The patients often feel a very distressing inability to call up the ideas they re-
quire. It seems that the individual ideas develop slowly and only under strong
stimulation ... usually nothing at all occurs to the patient, and he has to make
a special effort of will to continue his line of thought. Therefore, thinking is
laborious and slow, there is unreadiness in answering simple questions, lack of
comprehension, poverty of ideas.
Lewis also referred to a number of observers:
(i) Bleuler, who held that, in the inhibited melancholic, the retarda-
tion of thought and movement is always recognizable, cannot be
suddenly overcome and is not limited;
(ii) Ziehen, who stated that, in melancholia, depression and inhibition
of thought were both primary, although with reciprocal influence;
(iii) Gruhle, who stated that the Heidelberg research group "were firmly
convinced that true retardation (and especially its subjective re-
flection) was a characteristic of this affective disorder" (i.e., manic
depressive psychosis);
(iv) Wernicke, who gave "retardation of thought a predominant signifi-
cance in melancholia"; and
(v) Schneider, whose "vital depression" was held to be characterised by
"general inhibition and slowing of thought."
Lewis described "stupor" as an extreme form of inhibited action (or
immobility with an essential failure to react to exernal stimuli) but not
clearly involving any impairment of consciousness, and the close connec-
tion of stupor with retardation. He suggested three possible explana-
tions for thought inhibition present in retarded states: (1) interference
because of an "unpleasant mood-tone"; (2) interference because of pre-
occupation with affect-laden topics; and (3) a general disturbance due
to "profound, 'vital' changes in the organism, affecting the whole of
integrated psycho-biological activity."
Further considering "agitation," Lewis defined it as "the outward ex-
pression of a mental unrest, which may, however, be expressed only in
words" and held that the "inner unrest is the constant thing, the motor
unrest is variable." In Lewis' series, 16 patients at some time showed
marked motor restlessness: getting out of bed, wandering about, wring-
ing their hands, jerking their limbs or trunk, picking at their faces,
pulling off their bandages and so on. In some, there was concomitant
"retardation" or bewilderment, while others passed through a stage of
74 G. Parker and H. Brotchie

partial stupor. Lewis contrasted the speech patterns in retarded and


agitated patients, noting that when depressive talk is mainly reiterative,
the objective is apparently to importune. Compared to the bewildered
and the resigned, who tend to say little, he suggested that the agitated
speak a lot but without any constant association, and that their speech
content is even more marked by importunity.
The French psychiatrist Widlocher (1983) has also provided a com-
prehensive description of PMD, stating that "Retardation is considered
one of the major manifestations of depressive illness." He noted that it
is not clear whether the prefix "psychomotor" refers only to observable
behaviour or to "a general alteration of activity, including all mental
activities and experiences." Widlocher supported his observations by
historical references, including Darwin, who commented on "sad peo-
ple" who "no longer wish for action but remain motionless and passive,
or may occasionally rock themselves to and fro," and Dumas, who wrote
late in the nineteenth century that "melancholy as a mental entity does
not exist and, whatever its psychic or organic causes, motor phenom-
ena come before emotional ones, and melancholy is only perception of
physical condition."
Widlocher proceeded to challenge current views that mood change
was the primary disturbance in depression, and that retardation was
secondary to the mood state, to argue that "Retardation is a ... core
behavioral pattern, not merely a symptom ... [and the] primary distur-
bance in affective disorders."
After arguing that retardation had only been described in general
terms, or in a "global, impressionistic fashion" in the past, Widlocher
provided a rich behavioural description. He described, in part, the "gen-
eral slowing down of motor activity" with proximal parts of arms fixed,
trunk immobile, gait impaired, expression unchanging (with immobile
and blank face) and a paucity of movements of head and neck. Addition-
ally, he noted the slowing down of "psychic activity" with an inability
to respond quickly and spontaneously to the interviewer, responses brief
and laconic, difficulty in moving to a new theme, an impoverishment of
new themes, difficulty in concentration and often an apparent memory
disorder. He commented that speech is affected in both motor and cog-
nitive aspects, with the voice weak and monotonous, verbal flow slowed,
responses brief and the "perception of ruminative thinking." Widlocher
(1983) suggested that "retardation" was not the most appropriate term,
and that "freezing" and "clotting" might be more appropriate to de-
scribe the general slowing down.
Psychomotor Change: Historical Overview 75
Current classificatory systems offer variable status to PMD, generally
viewing it as reflection of depression severity but, in some instances,
as having diagnostic specificity. Thus, in DSM-IV (American Psychi-
atric Association 1994) one of the symptoms defining a "major de-
pressive episode" is observable psychomotor agitation or retardation
("not merely subjective feelings of restlessness or being slowed down"),
while a second-level diagnostic code of "melancholic features" includes
"marked psychomotor retardation or agitation" as one of several al-
lowed but non-mandatory contributing features. In the ICD-10 system
(World Health Organization 1992), "change in psychomotor activity,
with agitation or retardation" is one of several allowed to be chosen
from a list of features generating mild, moderate and severe "depressive
episodes." Such decision rules support our initial contention that PMD
is currently viewed principally as one of several indicators of depression
severity and perhaps having some potential as a sub-typing clinical fea-
ture (in being one of a set of non-obligatory features defining DSM-IV
melancholia).

Assessment of Psychomotor Change by Measures of


Depression
Despite its potential utility (whether as a feature of depressive severity,
or of type), PMD has tended to be assessed directly by very few items, if
any, in contemporary depression measures, although it must be conceded
that it may be indirectly assessed by items such as fatigue, anergia, loss
of concentration and anhedonia. The following overview suggests the
variable extent to which PMD is assessed by representative measures.
We note the focus on assessment by self-report information rather than
behavioural observation in the majority of measures.
The Hamilton Rating Scale for Depression (HRSD) (Hamilton 1960), is
essentially a clinician rating scale for assessing depression severity, with
2 of the 17 items assessing agitation (defined as "restlessness associ-
ated with anxiety") and retardation (ranging from slowness of thought,
speech and activity, through apathy to stupor). Factor analysis (Hamil-
ton and White 1959) suggested a first factor (defined by retardation,
depressed mood, guilt and loss of insight) corresponding to "the classi-
cal clinical description of Retarded Depression." Hamilton (1960) sub-
sequently noted that items loading on the second factor (somatic symp-
toms, insomnia, agitation, loss of weight, hypochondriasis and anxiety)
"vaguely" suggest "agitated depression." Subsequently, a more exten-
sive factor analysis (Hamilton 1967) emphasised that the scale assesses
76 G. Parker and H. Brotchie
"the severity of a patient's condition quantitatively ... [and] should not
be used as a diagnostic instrument."
There are two relevant concerns about the Hamilton scale in relation
to the assessment of PMD. First, while most definitions of agitation
focus on motor restlessness and make some attempt to distinguish it
from anxiety, the Hamilton definition probably blurs the boundary be-
tween anxiety and agitation. This is a major concern if, as we speculate,
anxiety may co-occur with any type of depression, while appropriately
defined motor agitation is specific to the melancholic depressive type.
Secondly, the rules for assessing agitation and retardation are unclear -
so that raters may variably rate on the basis of their own observation
and/or on their questioning of the patient about such features. Such a
concern is supported by an empirical study, where Schlegel et al. (1989)
failed to establish correlations between motor performance tests assess-
ing psychomotor retardation and retardation scores generated by either
the HRSD or the Bech-Rafaelson Melancholia Scale. Williams (1988) has
sought to redress any intrinsic lack of structure to the HRSD measure
by designing a "Structured Interview Guide." There, "retardation" and
"agitation" are based on interview observations, with the former being
assessed by noting "slowness of thought and speech; impaired ability to
concentrate; decreased motor activity" and the latter by features such
as fidgetiness, playing with hair, moving about and nail biting, although
again a number of the latter might well be tension-driven habits or anx-
iety behaviours rather than agitation as more strictly conceptualized.
While, as noted, Hamilton stated that his measure should be used
to assess severity only, a Hamilton Endogenomorphy Subscale was de-
veloped by Thase et al. (1983), using a subset of eight items from the
HRSD (including agitation and retardation), to provide a diagnosis of
endogenous depression, but comparison with DSM-m melancholic/non-
melancholic sub-groups established only a moderate overall agreement
rate.
Other clinician-rated depression scales vary considerably in their con-
sideration of PMD. Carney, Roth and Garside (1965) developed the
"Newcastle scales" to distinguish between endogenous and neurotic de-
pressed patients, and to predict ECT response, and each of the two
scales has 10 items. "Depressive psychomotor activity" is the only item
assessing PMD, is included in the diagnostic index only, and is denned
as describing any "objective evidence of psychomotor slowing, stupor
or agitation." Neither the Scandinavian-derived Diagnostic Melancholia
Psychomotor Change: Historical Overview 77
Scale (Bech et al. 1988) nor the Montgomery-Asberg Depression Rating
Scale (Montgomery and Asberg 1979) has any items assessing PMD.
The Raskin Three-Area Depression Scale is generally used as a global
assessment of three dimensions (behavioural signs, depressive symptoms
and verbal expressions of mood). The original study (Raskin et al.
1967) and a replication study (Raskin et al. 1969), both involving factor
analyses of interview, ward behaviour and self-report items assessing
clinical depressives, derived a "retardation in speech and behaviour"
factor (but from the ward behaviour ratings only), comprising items
such as expressionless face; flat, monotonous, slow speech; and slow
movement.
The Present State Examination or PSE (Wing, Cooper and Sartori-
ous 1974) is distinct in comparison to most other structured interview
schedules, in providing detailed definitions. PMD is likely to be assessed
by a number of items. For instance, the symptom "restless and fidget-
ing" is defined as "muscular tension expressed in motor activity" and
when severe "is expressed by pacing up and down, wandering about,
inability to sit still for very long." "Subjective anergia and retarda-
tion" is rated only on the "subject's subjective account" and not by
observing the subject's "speed of movement and response to examina-
tion." Behavioural ratings include "slowness and underactivity" (when
the "subject sits abnormally still or walks abnormally slowly or takes a
long time to initiate movement"), "agitation" (when there is "excessive
motor movement with a background of marker anxiety"), "slow speech"
(when there are long pauses before answering, and slowing or stopping
of speech altogether), "muteness," "restricted quantity of speech" and
"poverty of content of speech." The derived CATEGO classes include
three depressive categories (i.e., "depressive psychoses," "retarded de-
pression" and "neurotic depression"), but rules for class allocation tend
to require only a few of those features assessing PMD. Thus, a class diag-
nosis of "retarded depression" requires (a) depressed mood, and one of
the following three sets: (b) retardation, (c) guilt, self-depreciation, etc.,
or (d) agitation. While the authors report comparison data of CATEGO-
derived classes against clinical diagnoses obtained in two international
studies, validation of the CATEGO "retarded depression" class as a clini-
cal entity has not been established. PSE-derived data, however, do allow
the relationship between mood and PMD in depression to be examined.
For instance, Fleiss, Gurland and Cooper (1971) undertook a factor
analysis of items derived from a structured mental state interview in
one of the international studies, the US-UK Diagnostic Project cross-
78 G. Parker and H. Brotchie
national study. Thirteen items (including 10 from the PSE) generated
a "retarded speech" factor, 5 items (4 from the PSE) a "retarded move-
ment" factor, and 5 items (1 from the PSE) an "observed restlessness"
factor. The clinically diagnosed "psychotic depressives" were distin-
guished from the "neurotic depressives" in returning significantly greater
retarded speech and retarded movement, as well as observed restlessness
factor scores, but interestingly did not differ in severity on a 45-item de-
rived "depression factor" essentially assessing mood disturbance.
The Schedule for Affective Disorders and Schizophrenia was developed
by Endicott and Spitzer (1978), and contains 200 items, 30 describing
aspects of depression, collected from all available sources of information
to generate Research Diagnostic Criteria (RDC) depressive diagnoses for
current and/or previous episodes. Psychomotor agitation and retarda-
tion are assessed as 2 of the 30 items. Subsequent schedules, such as
the Diagnostic Interview Schedule (Robins et al. 1981) and the Struc-
tured Clinical Interview for DSM-m (Spitzer and Williams 1983) also
seek such information, as "psychomotor agitation or retardation" was
a DSM-III diagnostic criterion for major depression and for melancholia
when marked, and a DSM-III-R diagnostic criterion for major depres-
sion and melancholia. Similarly, psychomotor agitation or retardation is
one of the diagnostic criteria for the RDC "major depressive disorder,"
"endogenous major depressive disorder" and "minor depressive disor-
der" categories (Spitzer, Endicott and Robins 1980). In addition, the
RDC system allows "agitated" and "retarded" sub-categories of major
depressive disorder, with the former requiring two or more features such
as pacing, hand-wringing, inability to sit still and pulling of skin, while
the latter requires two or more of the following: slowed speech, increased
pauses before answering, low or monotonous speech, mute or markedly
decreased amounts of speech and slowed body movements.
Self-rating scales have, at best, few items directly assessing psychomo-
tor change. The 21-item Beck Depression Inventory (Beck et al. 1961),
the 13-item depression factor of the SCL-90 (Derogatis and Cleary 1977),
and the 20-item Center for Epidemiologic Studies Depression Scale (Rad-
loff 1977) have no such items, while the 20-item Zung Self-rating De-
pression Scale (Zung 1965) and the Leeds Scales for Self-Assessment of
Depression (Snaith, Bridge and Hamilton 1976) have only one item ("I
feel [am] restless and can't keep still") assessing psychomotor change.
The Carroll Rating Scale (Carroll et al. 1981b) was designed to mirror
the original items assessed in the Hamilton scale and, of the 52 items, 4
Psychomotor Change: Historical Overview 79
assess agitation and 4 retardation. Unfortunately, this seemingly useful
scale has not been used in many studies.
As part of a general review of the status and characteristics of PMD,
Cloitre, Katz and van Praag (1993) detailed videotape methodologies
and their clinical utility. Additionally, we note that attempts have been
made to measure bodily or facial expressions systematically, with, for
example, Pilowsky and Katsikitis (1994) describing a computerised pro-
cedure.

Tests of Psychomotor Functioning


While there have been numerous strategies used to measure behavioural
aspects of depression (Rehm 1987), most have generally examined be-
havioural correlates of a depressed mood rather than PMD. However,
as experiments assessing reaction times in depressives are reported in
Kraepelin's 1921 text (Berrios 1988), an extended period of interest in
measuring PMD is evident.
A detailed review by Yates (1973) covers tests of speed of responding,
reaction time, perseveration, work decrement and other constructs. A
paper by Popescu et al. (1991) illustrates some representative tests used
in applied studies, with those authors employing tests of visual reac-
tion time (i.e., pressing a key in simple and complex reaction time tests
variably involving single and multiple stimuli), of memory scanning, of
speech rate (time and speech pause time) and of motor speed by a "tap-
ping test," as well as the Stroop color-word test (naming colors of words
while undertaking a color-word interference task). A number of such
tests will be considered in later chapters detailing neuropsychological
studies.
Additionally, a useful overview has been provided by Greden and
Carroll (1981) of suggested strategies for measuring PMD physiologi-
cally (e.g., electromyographic activity in facial muscles, measurement
of speech phonation and pause times, movement-activated monitors of
motility), although few applied studies have been reported. For instance,
a simple measure was developed by Kupfer et al. (1972), involving an
electronic device that may be worn on a wrist or ankle, and a subse-
quent study (Weiss et al. 1974) reported significant correlations with
self-reported anxiety, but not with depression items. Royant-Parola
et al. (1986) undertook a similar study, with wrist motor activity being
continuously monitored during hospital stay for patients meeting DSM-
III major depression criteria (six considered to have endogenous and
80 G. Parker and H. Brotchie
four non-endogenous depression). As clinical improvement occurred,
activity levels increased, while "twenty-four hour immobility epochs"
decreased particularly during the night. The authors again failed to
find associations between activity levels and self-report data on mood
state. Diagnostic heterogeneity (i.e., non-melancholic as well as melan-
cholic depressives) may have influenced analyses - if our hypothesis (that
PMD is restricted to the melancholic depressive sub-type) is accepted.
Such findings have encouraged us to undertake similar "actimeter" stud-
ies - assessing (a) whether the motor disturbance precedes, accompa-
nies or follows the mood improvement as melancholia improves; (b) the
day and night pattern of motor activity in separate sub-groups of "re-
tarded" and "agitated" melancholic depressives; and (c) the extent to
which psychotropic medication may contribute to "immobility." Teicher
et al. (1989) have detailed a number of the methodological and statis-
tical issues involved in so assessing motor activity, noting the need for
powerful analytic techniques to provide meaningful summary statistics.
After reviewing a number of their own studies, they concluded that
their "present observations also emphasized the importance of separat-
ing psychomotor-retarded and psychomotor-agitated forms of depression
clinically" and that "the study of motor activity levels and rhythms
may . . . help in the delineation of depressive sub-types and in the de-
velopment of rational, innovative therapeutic strategies."
Apart from our CORE system (to be described shortly), we are aware
of only one other extended observation-based rating system of what
Popescu et al. (1991) termed "spontaneous" psychomotor activity (as
opposed to PMD elicited by test procedures) - the Depressive Retarda-
tion Rating Scale, developed by Widlocher (1983). That scale has 14
items with clear anchor points but, as suggested by its title, focusses on
retardation, and has no items assessing agitation.

Conclusion
In this review, we have attempted to demonstrate the varying extent to
which PMD has been described, interpreted and measured over time, to
make the point that current (largely symptom-based) depression mea-
sures and melancholia indices may not allow PMD's potential specificity
to melancholia to be established. In providing a brief historical overview
of PMD, we note that its ascendance as an observable and important
feature would appear to have occurred in the seventeenth century. It
subsequently appeared to lose its status as a core feature (particularly
Psychomotor Change: Historical Overview 81
in the English-language literature) early in the twentieth century, being
regarded more as a consequence of the mood state and therefore some-
what redundant in descriptive terms, while its potential specificity to
melancholia has not been recognised or conceded.
6
Development and Structure
of the CORE System
GORDON PARKER
DUSAN HADZI-PAVLOVIC

Endogenous depression, perhaps renamed, appears to be a useful concept for


both clinical and research purposes. It should be distinguished from non-
endogenous depression. (Andreasen, Grove and Maurer 1980)

Most psychiatric disorders are classified largely on the basis of their clinical
symptoms, or to be precise, a combination of the symptoms the patient com-
plains of and the behavioural abnormalities evident on examination. (Kendell
1983)

Introduction
Our University School of Psychiatry has had a lengthy investment in the
classification of depression, with the binary view favoured. The founda-
tion professor, Leslie Kiloh, had undertaken a number of classic studies
(e.g., Kiloh and Garside 1963) at Newcastle-on-Tyne, which made a
strong argument that "endogenous" and "neurotic" depression are in-
dependent entities. After moving to Sydney, he designed a replication
study (e.g., Kiloh et al. 1971), and follow-up of the cohort continues even
now (e.g., Andrews et al. 1990a). The orientation of the DSM-III classi-
fication (American Psychiatric Association 1980) to a more dimensional
view of the affective disorders was of concern to us as we judged that the
clinical and research utility in defining melancholia (or endogenous de-
pression) risked being lost, and we sought to address that potential lost
cause as our highest priority in setting initial objectives for our Mood
Disorders Unit (MDU). Such indigenous concerns were later shared by
North American researchers, with Zimmerman, Black and Coryell (1989)
noting that the DSM-III melancholia criteria had needed to be revised for
two primary reasons: first, because the criteria "did not predict treat-
ment response"; secondly, because "research suggested that the criteria

82
Development and Structure of the CORE System 83
did not identify a qualitatively distinct depressive subtype, but instead
differentiated patients solely along a severity dimension."
We therefore imagined a sequence of dispiriting steps emerging from
acceptance of the DSM-m definition:

(i) melancholia, as so defined, being unable to be validated, while at-


tempts at identifying aetiological mechanisms and at predicting
treatment response would fail as a consequence of the definition
(ii) disillusionment about melancholia leading to abandonment of the
concept and subsequent adoption of a dimensional model as the
default option
(iii) aetiological and treatment research (then underpinned by a dimen-
sional severity model) generating limited but quite variable findings
due to severity estimates being confounded by the mix of melan-
cholic and non-melancholic subjects in the particular sample being
studied.

As has been observed in another context, "Just as Michelangelo could


see his statues already formed within their blocks of marble - it is all
there, one merely has to set it free" (Ellard 1985), we had the belief that
melancholia is an entity, and that we should therefore attempt both its
clinical definition and distinction from non-melancholic depression, to
then allow re-examination of its aetiological determinants and biological
correlates. As a balance to enthusiasm, we had certain strategies to keep
hubris at bay. Specifically, we sought to:

(i) recruit very large samples of depressed patients;


(ii) ensure heterogeneity and adequate representation of putative sub-
types;
(iii) have a broad set of clinical variables;
(iv) have check mechanisms to ensure the accuracy of the data set; and
(v) focus on several recently introduced multivariate analytic strategies
which have the capacity to "group" patients rather than produce
"dimensions" of clinical features (as considered in Chapter 4).
In essence, we proposed what is commonly referred to gratuitously as
a "fishing expedition." Thus, we had no firm belief or initial hypothesis
that psychomotor disturbance (however defined or measured) is neces-
sary and sufficient to the definition of melancholia, and that it merely
required a superior measure or a better press in the eyes of the clinical
84 G. Parker and D. Hadzi-Pavlovic
research community. The first phase (here termed the CORE I study)
was exploratory and divergent. Large data sets were extensively anal-
ysed in a number of ways using traditional analytic approaches. The
most encouraging results, however, came following application of latent
class analysis (LCA) to our mental state item data set - while similar
analyses using symptoms were desultory. The second phase (the CORE
II study) involved a more convergent process - defining the nuances
of psychomotor disturbance, refining the measure, establishing its relia-
bility, examining its comparative properties, and examining the extent
to which it might provide a valid estimate of "melancholia." We now
overview the sequential CORE I and CORE II studies.

The CORE I Study


In addition to summarising material from our principal report (Parker
et al. 1990), we provide other unpublished information.

Samples and Diagnostic Options. We collected 202 consecutively


diagnosed primary depressed inpatients and outpatients from our ter-
tiary referral MDU sample, with each assessed according to a semi-
structured interview (Brodaty et al. 1987) by one of our consultant
researchers, and which allowed DSM-III, ICD-9, RDC and clinical diag-
noses to be generated. The last (psychotic, endogenous, neurotic and
reactive) were derived consensually by the consultants after an initial
interview, but could be subsequently modified after independent infor-
mation (e.g., from relatives or staff) and longitudinal review. As chronic
and treatment-resistant patients are over-represented at the MDU, we
recruited a (Sydney) sample of 262 depressed inpatients and outpatients,
using 30 Sydney-based clinicians (both private and public practitioners)
to rate patients on a similar set of symptoms and mental state signs,
as well as on several other measures detailed in the published report.
While analyses of both samples were described in our principal report,
we focus here on the MDU sample.

Clinical Features Assessed. As indicated in Chapter 12, we favoured


assessment of a comprehensive set of symptoms, essentially selecting
most of those that have been imputed as having relevance to melancho-
lia and psychotic depression over time. Thus, 26 potential endogeneity
symptoms were considered within the more detailed MDU data base
sample. In relation to mental state signs, it is important to note again
Development and Structure of the CORE System 85
that we did not seek merely to develop a measure of psychomotor dis-
turbance (PMD). Our objective was to try to capture the observable
domains of melancholia. To that end, in addition to rating a number
of presumed specific PMD behaviours, we assessed: (a) general attitude
to the interviewer; (b) severity of mood (affect); (c) guilt and nihilism
- as expressed verbally and non-verbally at interview; and (d) endogen-
ous quality (but without any description), as some clinicians have held
that melancholic patients have a characteristic mien. Each item had
four anchor points, with one pole representing no disturbance and the
remainder quantifying increasing levels of clinician-judged disturbance.

Statistical Analyses. Respecting many previous inquiries of this na-


ture, we first used principal components analysis (PCA), which has the
capacity to identify sets of items that group together but which cannot
directly address groupings of patients. Secondly, we used latent class
analysis (LCA), a procedure capable of identifying indicators separat-
ing members and non-members of a presumed latent or underlying class
(here melancholia) and which has been described in Chapter 4. While
a number of LCA methods have been proposed, we used the taxometric
model of Golden (1982). We examined for evidence of discontinuity (and
especially for any support for the binary hypothesis of two groups) by
applying mixture analyses to total scores on the subset of items identi-
fied by these two approaches. The mixture model was estimated using
the method of Hasselblad (1966) and the number of component sub-
populations tested using the likelihood ratio test (Everitt 1981).

Characteristics of the MDU Sample. In this, the key sample fo-


cussed on in this overview:
(i) Clinical diagnosis assigned
(a) 12% to psychotic depression
(b) 41% to endogenous depression
(c) 40% to neurotic depression
(d) 7% to reactive depression
(ii) DSM-III diagnosis assigned
(a) 8% as having a major depressive episode (MDE) with psychotic
features
(b) 30% as MDE with melancholia
(c) 53% as MDE alone
86 G. Parker and D. Hadzi-Pavlovic
(d) 7% as having dysthymia
(e) 2% as having other depressive diagnoses
(iii) The Newcastle Scale (Garney, Roth and Garside 1965) assigned
(a) 20% as having endogenous depression
(b) 80% as having non-endogenous depression
The mean age was 43 years; there was a female preponderance (69%),
the mean episode duration was 81 weeks and the mean Zung depression
score was 55.0. By contrast, the Sydney sample was somewhat older,
had a briefer mean episode duration, and had a higher rate (61%) of
clinically diagnosed psychotic and endogenous depressive patients. As
the status of psychotic depression is unclear, but is often viewed as
synonymous with melancholia (see Chapter 12), we consolidate psychotic
and endogenous depressives in a number of analyses.

Examination of Symptoms. Table 6.1 first notes the frequency of dif-


fering symptoms reported by those allocated as having a psychotic/endo-
genous versus neurotic/reactive depression, with odds ratio (OR) show-
ing that only 13 (50%) of the symptoms were significantly over-re-
presented in the first group. Equally importantly, it can be readily
observed that few symptoms (apart from delusions) showed specificity
to the first group, an issue that we will return to in expressing concerns
about symptom-based measures of melancholia.

Creating an Endogeneity Symptom Scale. We sought to develop a


refined endogeneity symptom scale (SYMPTOM). Importantly, our LCA
failed to produce a refined set of symptoms that were either clinically
or historically compatible with the definition of endogenous depression
or melancholia. When we undertook a PC A, the first two components
accounted for 14% and 8% of the variance. The first component (defined
by those 15 items loading > 0.30) was clearly compatible (Table 6.1)
with published factorial definitions of endogeneity or melancholia and,
in essence, suggested a retardation dimension, with anhedonia, non-
reactivity, appetite and weight loss, together with slowed and indecisive
thoughts dominating. The second (more clearly bipolar) component was
less readily interpret able.
We created a SYMPTOM score in the MDU sample by summing scores
on the 14 symptoms loading highest on the first factor (Table 6.1). The
next highest loading item (delusions) was excluded due to the low preva-
lence. The mean SYMPTOM score was 8.2, and was independent of sex,
Development and Structure of the CORE System 87

Table 6.1. Analyses of symptoms in CORE I study: their capacity to


differentiate clinical diagnoses (with odds ratios quantifying the greater
chance of a symptom being present in the psychotic/endogenous
depressive groups), and pre-rotation loadings on the first two factors
extracted by the PC A

Percentage Factor
affirming Odds loadings
Symptom PD/ED ND/RDf ratio I II

Consummatory anhedonia 82% 71% 1.86 0.65 0.08


Anticipatory anhedonia 79% 65% 2.02* 0.63 0.35
Appetite loss 51% 21% 3.89* 0.58 -0.33
Weight loss 39% 26% 1.82 0.52 -0.32
Non-reactivity
(a) in social situations 74% 67% 1.40 0.54 0.41
(b) to likely pleasant event 75% 67% 1.52 0.47 0.27
Indecisiveness 83% 53% 4.51* 0.47 0.10
Preoccupied with
unpleasant thoughts 58% 52% 1.27 0.44 -0.05
Slowed thoughts 87% 69% 3.01* 0.44 0.25
Terminal insomnia 51% 31% 2.40* 0.43 -0.26
Non-variable mood 57% 43% 1.82* 0.42 0.10
Loss of interest and pleasure 81% 70% 1.82 0.39 0.23
Constipation 47% 29% 2.16* 0.33 0.07
Retardation 71% 45% 2.95* 0.33 0.30
Delusions 16% 2% 9.32* 0.32 -0.21
Distinct mood 93% 72% 5.18* 0.25 0.15
Agitation 47% 43% 1.19 0.23 -0.26
Middle insomnia 35% 38% 0.88 0.15 -0.23
Suicidal thoughts 54% 66% 0.58 0.12 -0.01
Initial insomnia 35% 32% 1.14 0.12 -0.09
No energy in morning 44% 27% 2.12* -0.01 0.32
Anger with others 18% 44% 0.28 -0.04 -0.01
Mood worse in morning 56% 41% 1.90* -0.05 0.29
Hypersomnia 19% 26% 0.67 -0.19 0.55
Weight gain 16% 19% 0.81 -0.20 0.59
Increased appetite 18% 28% 0.57 -0.32 0.54
f Clinical diagnoses of depression: PD == psychotic; ED = endogenous; ND =
neurotic; RD = reactive.
* Significant at P < 0.05.
88 G. Parker and D. Hadzi-Pavlovic

age and duration of depression, but was associated with severity as as-
sessed by the self-report Zung (r = 0.27) and GHQ (r = 0.22) measures,
as well as with the clinician-rated Hamilton (r = 0.35) and Newcas-
tle (r = 0.40) measures. Psychotic/endogenous depressives, as defined
clinically, by DSM-ni or by the RDC system, returned higher SYMPTOM
scores than the neurotic/reactive depressives.

Refining Signs and Creating a CORE Score Sample. We pursued


strategies similar to those employed for the symptoms. In the PCA, the
first unrotated component accounted for 41% of the variance. Table 6.2
lists the loadings on this first component which (and as confirmed in
the Sydney sample) strongly defined a retardation dimension. Oblique
rotation of varying numbers of components consistently retained this
dimension, while the next four factors each accounted for only small
percentages of the variance.
Table 6.2 also reports results from the LCA. As noted earlier, the
basis of an LCA is that, in a sample assumed to comprise more than
one class, the classes (here melancholies and non-melancholies) can be
distinguished by different patterns of response on an item set. For each
class, the LCA estimates the probability of a given response on a given
item for a member of that class (item probabilities). If items are able to
differentiate, item probabilities should show a large difference (i.e., some
specificity) across the two classes; Table 6.2 demonstrates differential
loadings ranging from 0.53 (for the dull/inattentive item) to 0.75 (for
the reduced spontaneous talk item) for the items retained in the LCA.
If an item is specific to melancholia, we would wish to see a probability
representing its prevalence in the melancholic class and a zero probability
in the non-melancholic class. A high probability in the melancholic
class is not mandatory, unless the feature is a necessary condition, as
even a feature specific to the class (e.g., delusions) may still be rare.
For the retained items, a number had loadings of less than 0.10 (e.g.,
immobility; slowed movements; inaudible) in the Class B (or putative
non-melancholic class) indicating their virtual absence in that class, and
quite high loadings in Class A (putative melancholia), so demonstrating
desired specificity.
Thirteen items were retained in the final set in the MDU sample (as
opposed to 14 in the Sydney sample), and with 9 identified consistently
in both samples. Inspection of the retained items indicated that, as for
the PCA, the LCA had defined a retardation domain - evident from
inspection of the highest loading items in the final set and by failure to
Development and Structure of the CORE System 89

Table 6.2. Analyses of mental state signs in CORE I study:


pre-rotation first principal component and probability estimates
(using LCA) for the final 13 items retained in the MDU sample

Item probabilities
Factor in LCA
Mental state sign loading! Class A Class B

Reduced spontaneous talk 0.79 0.86 0.11


Fixed immobile face 0.78
Unresponsive to interviewer 0.77 0.86 0.16
Poverty of associations 0.76 0.94 0.25
Slowed movements 0.75 0.73 0.07
Immobility 0.74 0.78 0.06
Dull/inattentive 0.73 0.59 0.06
Mute or reduced speech 0.71 0.71 0.05
Reduced themes 0.71 0.81 0.22
Depressed affect maintained 0.70
Persistently miserable 0.70
Slumped posture 0.70 0.76 0.15
Slowed speech 0.70 0.72 0.06
Depressed affect initially 0.69
Inaudible 0.67 0.68 0.04
Self-preoccupied 0.67 0.90 0.23
Not cheered by interviewer 0.67
Retardation of movement 0.66 0.74 0.12
Speech favouring self 0.65
Endogenous quality 0.63
Impaired concentration 0.59
Persistently worried 0.57
Nihilistic 0.57
Impaired insight 0.50
Uncooperative attitude 0.46
Agitated movement 0.41
Facial agitation 0.39
Observable anxiety 0.33
Excessive guilt 0.32
Querulous and irritable 0.30

Class prevalence 33%.O 67%.O

f Factor loadings on first factor from principal components.


90 G. Parker and D. Hadzi-Pavlovic
include the two putative agitation items (i.e., facial agitation, agitated
movement) or the observable anxiety item.
The LCA procedure also allows the probability of class membership
to be calculated for any individual using his or her pattern of responses.
Table 6.2 shows that the LCA allocated 33% of the MDU sample to
Class A (putative melancholia), an identical percentage (but not neces-
sarily the same subjects) as allocated a clinical diagnosis of psychotic or
endogenous depression.
We then finalised our set of CORE I signs to those 15 items suggested
most consistently across the two multivariate techniques and across the
two samples. Table 6.3 examines the percentage of subjects in each sam-
ple rated as showing some abnormality (i.e., scoring 2-4) on each item.
Those diagnosed clinically as having psychotic/endogenous depression
were more likely to rate positively on most items, as quantified by odds
ratio (OR) estimates.
We then computed a CORE score for each subject by summing raw
(1-4) scores on each of those 15 items. Mean CORE scores were 24.2 (SD
7.8) in the MDU sample (males 23.3, females 24.6), and 27.8 (SD 8.5)
in the Sydney sample (males 27.6, females 27.9).

Interpreting CORE I Scores. We undertook mixture analyses to


determine the number of subpopulations reflected in CORE scores, with
those analyses indicating three populations in each sample and with very
similar means (and standard deviations) across the two samples. Com-
parisons against clinical notes and formal diagnostic decisions suggested
that the bulk of the clinically diagnosed neurotic/reactive depressives be-
longed to the first population (where the means indicated that subjects
scored minimally on the CORE items), while most of the clinically di-
agnosed psychotic/endogenous depressives belonged to the second and
third populations; the non-psychotic melancholic patients were more
likely to be allocated to the second, and the psychotic melancholic pa-
tients were more likely to be allocated to the third population.
We undertook a number of analyses (see Parker et al. 1990) exam-
ining for relationships between CORE scores and other variables. In
summary, CORE scores increased with age and with older age of onset of
initial depressive episode, but were independent of sex and of duration
of depressive episode. CORE scores were associated with both Newcas-
tle (r = 0.52) and Hamilton (r — 0.42) scores (observer-rated measures
requiring a judgment on psychomotor disturbance), but not with Zung
(r = 0.04) or General Health Questionnaire (r = 0.06) scores (self-rated
Development and Structure of the CORE System 91

Table 6.3. CORE I study: percentage of subjects rated as having some


abnormality on the final 15 CORE items, by clinical diagnosis, and the
extent to which such a sign was more likely to be present in those with
a psychotic/endogenous depression

Clinical diagnosis
OHrlc
V/tlClb
CORE I items PD/ED ND/RDf ratio
Unresponsive to interviewer 54 30 2.72***
Dull/inattentive 36 10 5.04***
Fixed immobile face 77 54 2.95***
Self-preoccupied 55 41 1.75
Inability to be cheered by interviewer 86 75 2.04
Slumped posture 44 26 2.23**
Immobility 38 18 2.78**
Slowed movements 38 15 3.61***
Slowed speech 35 16 2.82**
Mute or reduced speech 35 16 2.82**
Poverty of associations 61 42 2.06**
Impaired insight 30 16 2.25*
Nihilistic 34 19 2.19*
Observable anxiety 78 74 1.25
Endogenous quality 75 31 6.87***
t Clinical diagnosis of PD = psychotic; ED = endogenous; ND = neurotic;
RD = reactive.
*** p < 0.001; ** p < 0.01; * p < 0.05.

measures of the severity of depression and psychological dysfunction re-


spectively).
We examined mean CORE scores for subjects in each sample and diag-
nosed according to different diagnostic systems: DSM-III and CLINICAL
diagnoses in each sample and, additionally in the MDU sample, RDC
diagnoses and Newcastle assignment. In each sample, and for each diag-
nostic system, those diagnosed as having psychotic/endogenous/melan-
cholic depression scored significantly higher than the remaining depres-
sives, while in sub-analyses, those with psychotic depression returned
higher CORE scores than those with endogenous/melancholic depression.
Differentiation of clinical sub-types by CORE scores was considerably
more evident than shown for similar analyses of symptom scores.

Could a CORE I Cut-off Score Distinguish Melancholia? Judg-


ing that melancholic depressives were most likely to be in the second and
92 G. Parker and D. Hadzi-Pavlovic
third groups defined by our mixture analyses, we sought to determine
a CORE cut-off for discriminating "melancholia" by examining receiver
operating characteristic (ROC) curves (Murphy, Berwick and Weinstein
1987), with those curves plotting false against true positives, and with
points on the curve representing the changing performance of the diag-
nostic test. In the MDU sample, the most discriminating cut-off CORE
score was 20/21, while in the Sydney sample it was 24/25. Of those
clinically diagnosed as having psychotic or endogenous depression, 78%
in the MDU and 84% in the Sydney sample scored above the derived
cut-off points. In addition, we had some dexamethasone suppression test
(DST) data for 29 patients in the MDU sample, so allowing potential
validation against a "biological index" of melancholia. Non-suppression
(using a cut-off of 138 nmol/1) occurred in 45% of those exceeding the
high CORE cut-off (of 21 or more) as opposed to a zero non-suppression
rate in those scoring below that cut-off. Further validity studies (i.e.,
naturalistic outcome, neurocognitive tests, prediction of ECT and an-
tidepressant drug response) will be considered in Chapters 8-10.

CORE I and SYMPTOM Score Associations. The overall cor-


relation between total CORE and SYMPTOM scores was significant (r =
0.39, P < 0.001), equivalent to a shared variance of approximately 16%.
When individual symptoms were examined, those most strongly linked
with CORE scores were: preoccupation with unpleasant thoughts, indeci-
siveness, non-variable mood, constipation and retardation. Significantly,
but less clearly associated were: marked appetite loss, consummatory
anhedonia, anticipatory anhedonia, non-reactivity in social situations
and slowed thoughts, with significant associations suggesting that these
symptoms again denned a retardation dimension.

CORE and SYMPTOM Scores as Comparative Predictors of


Clinical Diagnosis. While recognising the risk of clinical diagnoses
reflecting our own biases, we also viewed them as having advantages
(in representing a judgment made by experienced clinicians on both
cross-sectional and longitudinal information). We therefore undertook
a discriminant function analysis to examine the comparative prediction
of clinical diagnostic assignment to melancholia versus non-melancholia
(here psychotic/endogenous vs. neurotic/reactive) by total SYMPTOM
and total CORE scores, and established canonical correlation coefficients
of 0.66 and 0.59 respectively, suggesting that symptom and sign scores
had a similar ability to predict clinical diagnosis, and with the overall
Development and Structure of the CORE System 93
classification rate being 68%. Aware that our capacity to evaluate pre-
dictability was confounded by doubtful and mixed diagnoses (and by
some subjects not being at episode nadir), we undertook sub-analyses
of more clinically distinct groups.
Specifically, we compared (a) 26 subjects who were rated as (i) definite
RDC psychotic or endogenous depression and (ii) having a DSM-HI-R
major depression with melancholia or delusions and (iii) endogenous on
the Newcastle Scale, versus (b) 47 subjects who (i) had definite RDC
major depression and (ii) had a DSM-III major depression or adjustment
disorder and (iii) were rated as non-endogenous on the Newcastle Scale.
A discriminant function analysis, entering each of the 15 CORE signs
and each of the 14 symptoms, had an overall correct classification of
92% (and a distinct bimodal distribution). More importantly, the entry
of CORE signs alone reduced Wilks' lambda to 0.39 (F = 5.87, equivalent
to R2 = 0.61), whereas the entry of symptoms alone reduced lambda to
0.55 (F = 3.40, R2 = 0.45), suggesting that, overall, the CORE score
was a stronger discriminator than the SYMPTOM score. This conclusion
was confirmed by varying the entry order of predictors and, in effect,
controlling for shared variance. Here, after entering the CORE score
first, the entry of the SYMPTOM score reduced Wilks' lambda further
by 0.18. By contrast, the entry of the CORE score (after entering the
SYMPTOM score first) reduced lambda further by 0.33. These analyses
suggest that, in clinically distinct subgroups (with diagnostically difficult
or mixed-diagnosis patients excluded), our refined sign-based measure
was a stronger predictor than the refined endogeneity symptom score
and that, after shared variance had been accounted for, the CORE score
made an independent contribution distinctly greater than that made by
the SYMPTOM score.

Clinical Validity of the CORE I Score


The synonym for melancholia, endogenous depression, captures the view
that melancholia is more biological in aetiology and largely independent
of the subject's social and personal world. Applying the aetiological
view of melancholia to psychosocial variables, Zimmerman et al. (1986b)
argued that melancholic depressives ("endogenous depressives") should
be less likely to have a pre-morbid personality disorder and less likely
to misconstrue and react to neutral and negative events; should have
a lower frequency of stressful life events in the preceding year; should
have a lower frequency of marital separation and divorce, and should
94 G. Parker and D. Hadzi-Pavlovic
have better social supports. If valid, then a diagnostic system defining
melancholic and non-melancholic depressives should demonstrate dif-
ferential relevance of such psychosocial variables to defined depressive
sub-types.
Zimmerman and colleagues then compared the capacity of a number
of diagnostic systems to discriminate on those five broad psychosocial
variables. They reported that the Newcastle index was the most discrim-
inating system (positive on pre-morbid personality disorder, frequency
of stressful life events, and separation-divorce), whereas DSM-III was val-
idated by only one variable (frequency of stressful events), while the RDC
and Feinberg and Carroll systems failed to differentiate.
We undertook a similar comparative analysis, examining the capacity
of the following systems to differentiate:

(i)those who scored above the CORE cut-off (HIGH CORE);


(ii)those who scored above the SYMPTOM cut-off (HIGH SYMPTOM);
(iii)those meeting DSM-III melancholic criteria (DSM);
(iv) those meeting RDC psychotic or RDC definite endogenous depression
criteria (RDC);
(v) those receiving a clinical diagnosis of psychotic or endogenous de-
pression (CLINICAL); and
(vi) those who scored 5 or more on the Newcastle Scale (NEWCASTLE).
The potential validators included a number of the clinical givens of
melancholia (e.g., younger age; younger age at first episode; briefer
episodes; less severe depression) as well as several psychosocial valida-
tors. The last included

(i) Parental Bonding Instrument scale scores (with a number of studies


having demonstrated that non-melancholic - but not melancholic -
depressed subjects are more likely to report their parents as uncar-
ing and/or overprotective, domains measured by the instrument);
(ii) a range of deprivational parental experiences;
(iii) Intimate Bonding Instrument scale scores (reflecting a number of
studies demonstrating that non-melancholic depressed subjects were
more likely to report their partners as less caring and as more con-
trolling, domains reflected in the measure);
(iv) social functioning assessment;
(v) several personality measures of neuroticism and interpersonal sen-
sitivity; and
Development and Structure of the CORE System 95
(vi) estimates of acute and chronic stressors.
We now summarise findings from that study (Parker et al. 1991d).
Those who scored as HIGH CORE (compared to remaining subjects) were
found to have the following characteristics:
- were older at onset of first episode;
- experienced briefer episodes in the past; and
- had a more severe episode as rated on the Newcastle and Hamilton
scales
They were were also significantly less likely to report:
- either parent as uncaring;
- several parental deprivational experiences (i.e., dysfunctional fa-
thering, dysfunctional parental marriage, parental separation);
- their partner as uncaring;
- a recent break-up in an intimate relationship; and
- exposure to chronic stressors
To the extent that the examined variables validate melancholia in
terms of clinical characteristics and risk variables, the CORE scale as-
signment was intermediate if merely the number of significant variables
are considered, being less successful than CLINICAL and NEWCASTLE,
equivalent to DSM, and superior to RDC and high SYMPTOM assignment.
However, as those two most successful systems allow the clinician rater
to incorporate a number of judgments about personality and life event
precipitants, as well as historical course and treatment response, the
comparative success of assignment by the CORE measure (as a cross-
sectional behavioural snapshot) appeared quite impressive.

Discussion of the CORE I Study


In our attempt to differentiate those with melancholia from a residual
group, our analyses suggested the superiority of assessing mental state
signs as opposed to the historical focus on symptom-based approaches.
In the absence of "gold-standard" validating criteria, delineation of
depressive types is limited but generally requires a "bottom-up" strat-
egy - with the objective being to isolate allied categorical or dimensional
characteristics. Most attempts to derive typologies have proceeded in
this way (using multivariate techniques such as factor and cluster analy-
sis), but the strategy is limited by the difficulties in validating the set of
96 G. Parker and D. Hadzi-Pavlovic

derived features (either against other measures of melancholia or against


the clinical givens of melancholia, or by other validation strategies).
Using a bottom-up approach on a large data set, a clear retardation
dimension or construct was evident in both our refined set of signs and
symptoms. As noted in previous chapters, a number of factor analytic
studies have demonstrated the central importance of retardation. Again,
it is worthwhile to quote from the review by Nelson and Charney (1981):
"Psychomotor change is the symptom most strongly associated with au-
tonomous depression. Retardation was especially consistent in the factor
analytic and drug response studies," while retardation again "received
strong support" in their review of cluster analytic studies.
In our initial report (Parker et al. 1990) we argued for the possible
utility of a sign-based approach to the definition of melancholia on two
broad grounds. First, our empirical analyses suggested that CORE scores
(and the derived high CORE sub-group) had stronger links to variably
defined psychotic/endogenous/melancholic categories than did equiva-
lent refined SYMPTOM scores. Additionally, we undertook a number of
validation studies (overviewed here and in Chapters 8-10) which again
favoured the refined signs as a measure of melancholia. Secondly, we
articulated a number of methodological concerns about any diagnostic
system weighted to symptoms (e.g., reliance on subjects' self-reports,
particularly those with severe and/or psychotic depression; difficulties
in determining whether to rate at nadir or at the time of assessment;
medication effects), in addition to expressing doubts about the validity of
some being true endogeneity symptoms. Thus, if melancholia is able to
be defined by a domain of psychomotor retardation, we suggested that a
reliable observational measure might provide a more valid estimate than
reliance on symptom reports.

The CORE II Study


The preliminary CORE I findings were clearly encouraging and sug-
gested a more hypothesis-testing second study, one shaped by a number
of provisional findings. Instead of merely developing and administering
a rather general mental state examination of various signs, we then set
out to develop a refined measure focussing on psychomotor disturbance.
The CORE I study had suggested the utility of assessing psychomotor
retardation but had not identified any utility to psychomotor agitation.
It remained unclear whether this reflected a lack of agitation items in our
initial study (as well as a lack of item definition), or the fact that agita-
Development and Structure of the CORE System 97
tion was not relevant. We therefore decided to examine the comparative
properties of a refined mental state examination, focussing now on signs
of psychomotor disturbance (both retardation and agitation signs) and
thus undertook the CORE II study.
In redesigning the mental state examination, the priority, then, was
to redress several limitations suggested or identified for the CORE I
items. First, that system lacked detailed descriptors, operational defi-
nitions and clear anchor points. Secondly, some items such as endog-
enous quality invited or allowed non-behavioural judgements. Thirdly,
there were only two items addressing agitation. Fourthly, the lack of
operational definitions and distinctions led to the risk that anxiety and
agitation (being undefined) were not sufficiently well differentiated.

CORE II study: Sample and Assessment Procedures. Patients


were enrolled (Parker et al. 1994) subject to a primary clinical diagnosis
of a depressive disorder present for at least two weeks, with 185 recruited
from our tertiary referral MDU. An additional 228 were assessed by us
at several other Sydney hospitals ensuring sample heterogeneity. The
mean age of the 413 was 51.0 (SD 18.2) years; 273 (66%) were female;
and 50% (205/413) were inpatients. Subjects were randomly assigned
to two sub-samples for the principal multivariate analyses: a "training"
sample of 250 patients (adequate in size as a development sample for
the multivariate analyses) and a "validation" sample of the remaining
163 subjects to allow validation checks on solutions generated within the
training sample.
All patients completed the Zung (1965) depression scale, with the
mean score being 56.2 (SD 8.8). Those in the MDU sample were also
rated on the 21-item Hamilton (1960) Scale, with the mean score being
20.0 (SD 7.3).

CORE II Diagnostic Allocations. The psychiatrist raters were re-


quired to choose a principal clinical diagnosis (i.e., psychotic [PD]; en-
dogenous [ED], neurotic [ND], and reactive depression [RD]) according
to brief definitions. Thus, subjects were allocated to the ED diagnostic
category if there was evidence of classical features of endogeneity (e.g.,
significant psychomotor disturbance, vegetative features, pervasive an-
hedonia and non-reactive mood) and absence of delusions and halluci-
nations. If there was (in addition) clear evidence of delusions and/or
hallucinations, subjects were allocated to the PD category. A diagnosis
of ND or RD required no or few classical endogeneity features, with ND
98 G. Parker and D. Hadzi-Pavlovic

also requiring supportive evidence of a pre-morbid neurotic personality


style. By contrast, RD was generally diagnosed for patients without
any evidence of a neurotic personality style and who had developed de-
pression following a significant life event, and included bereavement and
related loss responses.
The final clinical diagnoses were: 73 (18%) PDs, 140 (34%) EDs, 158
(38%) NDs and 42 (10%) RDs. As there were few RDs, and as both ND
and RD are viewed as non-melancholic disorders, we aggregated these as
a combined ND/RD group in many analyses. Similarly, we aggregated
the PD and ED sub-groups in some analyses, although we recognise that
it remains contentious whether psychotic depression is a separate type
or a more severe expression of melancholia.
DSM-III-R diagnoses and Newcastle (Carney et al. 1965) scores were
available for the MDU patients. A DSM-III-R diagnostic algorithm estab-
lished that 180 (97%) met criteria for major depressive episode (MDE),
with 105 of these (57% of that sub-sample) meeting additional criteria
for melancholia or for having psychotic features. All but two of the 33
(18%) who met criteria for dysthymia also met higher-order MDE cri-
teria. The cut-off of 6 or more was used to assign MDU sub-sample
members to endogenous depression on the Newcastle Scale, with 59/185
(32%) being so assigned.

CORE II Clinical Symptoms and Signs. We selected 18 potential


endogeneity symptoms, either on the basis of their being previously sug-
gested from multivariate analyses (see Chapter 12) or by their inclusion
in diagnostic systems such as DSM-III-R (American Psychiatric Associ-
ation 1987), but we did not include formal psychotic features such as
delusions and hallucinations in these refining analyses. During a semi-
structured clinical interview, symptom data were coded on four-point
scales measuring absence (0) or, if present, severity (1, 2 or 3).
Twenty-four provisional signs of psychomotor disturbance were gen-
erated - 14 from the CORE I item set, and the remainder focussing on
psychomotor agitation and motor delay. Rating instructions and opera-
tional definitions of items were written and modified during the study to
resolve ambiguities. A rating of 0 indicated the "absence or triviality"
of a feature, while ratings of 1, 2 or 3 indicated that the feature was
distinct and pathological, and (respectively) slight, moderate or severe.
(Chapter 14 further discusses the rating of signs.)
Development and Structure of the CORE System 99
Other CORE II Data. To distil a possible validating biological index,
we sought information on background and treatment issues previously
suggested by Zimmerman et al. (1989) as over-represented in those with
melancholia, including a family history of depression; a previous episode
of depression or mania; a history of distinct remission of any previous
depressive episode (denned as an improvement of "75% or more"), both
generally and in response to an antidepressant and/or to ECT; and the
presence of delusions and of hallucinations.
In the second recruitment year, we commenced two studies in the
MDU sub-sample for independent validation of the sign-based system.
We report the extended dexamethasone suppression test (DST) study
in Chapter 8, and reaction time study in Chapter 9; however, as an
initial DST analysis involving 62 inpatients is considered in this chapter,
we note the DST study methodology. Pre-dexamethasone serum was
collected for cortisol assay at 1600 hours (day 1), with dexamethasone
1 mg given orally at 2300, and with further serum samples for cortisol
collected at 0800, 1600 and 2300 (day 2). Non-suppression was defined
as a cortisol value above 137 nmol/1.
As a check on our data set of clinical features, we sought patients to
nominate a relative who would be prepared to complete a questionnaire
independently, with those witnesses being required to rate the patient
"over the last month for when he/she was most depressed." This study
component has been described in detail (Parker et al. 1992a); in essence,
however, symptoms were generally assessed and rated according to the
same procedures used by the clinicians. For signs, we felt obliged to
provide simple clinical descriptions for some features, with the level of
correspondence between clinician rater and witness rater forms varying
from slight to moderate in relation to both the descriptors and the anchor
points.
Note: As considerable material is now reviewed, we provide summaries
at the end of some sections rather than leave such material to the dis-
cussion.

CORE II Study: Refining the Signs. We now summarise material


previously reported more extensively (Parker et al. 1994). Three signs
were deleted from the original set of 24 because of progressive doubts
by the consultant raters that they were truly signs. Additionally, one
item (depressed affect) was discarded, as positive ratings were virtually
ubiquitous across the sample (97%), and another (self-preoccupation)
because of difficulties in its objective definition.
100 G. Parker and D. Hadzi-Pavlovic
Our observable anxiety item (as manifested by features such as tremor,
hyperventilation, sweating and muscular tension) demonstrated unan-
ticipated properties. Firstly, scores on that sign correlated minimally
with most other signs, including our chosen agitation items. Specifi-
cally, it had positive (but weak) polychoric correlations only with the
facial apprehension (0.12), facial agitation (0.25) and motor agitation
(0.15) items. Secondly, in a screening PC A, anxiety was the only item
to make a significant negative loading (—0.52) on the first component
(where all other items loaded positively), and its loading on the third
(agitated) component was trivial (0.01) in comparison to loadings of the
other agitation items (e.g., facial agitation = 0.78; facial apprehension =
0.78; motor agitation = 0.74) on that component. Whether as a conse-
quence of our operational definitions or of its phenomenological status,
the anxiety item appeared independent of the agitation items, and was
therefore deleted.
Table 6.4 shows that the remaining 18 signs were rated as present
in 12-61% (mean frequency = 32%) of the subjects, while comparative
analyses established that each of the 18 items was significantly more
likely to be present for those diagnosed clinically as having PD or ED
(compared to ND or RD), with ORs (quantifying the greater likelihood
of PD/ED diagnoses if that sign is present) ranging from 4 to 104, and
with a median OR of 14.

Summary. An initial sign item, observable anxiety, appeared indepen-


dent of a set of agitation items and was deleted as being a poor indicator
of psychomotor disturbance. Each of the final 18 signs was distinctly
over-represented in the clinically diagnosed psychotic and endogenous
depressed patients.

The Structure Underlying the Signs


Maximum likelihood factor analysis was used to identify dimensions or
sets of items that grouped together. The retained factors were rotated
obliquely (Promax) to produce a pattern matrix where each variable had
non-zero loadings on as few variables as possible (McDonald 1985). We
restricted our attention to two-factor and three-factor models for two
reasons. First, the first three factors accounted for 52%, 15% and 5% of
the variance of the original matrix, suggesting that central dimensions
were contained in the first few factors. Secondly, extraction and rotation
of a large number of factors led to unstable solutions. Before rotation,
Development and Structure of the CORE System 101

Table 6.4. CORE II study: percentage of patients with signs and


symptoms rated as present in total sample and in diagnostic sub-groups

Percentage affirming feature


All PD ED ND/RDf Odds
Clinical feature n = 413 73 140 200 ratioj
SIGNS
Non-interactiveness 37 73 52 14 8.9***
Facial immobility 52 78 78 25 10.9***
Postural slumping 42 69 56 22 5.4***
Non-reactivity 61 96 84 33 14.6***
Facial apprehension 60 89 71 42 4.7***
Delay in responding verbally 27 64 39 4 22.1***
Shortened verbal responses 32 64 50 8 15.0***
Inattentiveness 23 59 32 4 16.9***
Facial agitation 28 63 28 15 3.9***
Body immobility 37 63 61 11 13.2***
Motor agitation 30 60 43 10 9.1***
Poverty of associations 40 84 61 10 20.8***
Slowed speed of movement 32 59 52 9 12.1***
Verbal stereotypy 24 63 32 4 20.6***
Delay in motor activity 18 44 29 1 103.8***
Impaired spontaneity of talk 37 64 59 11 13.4***
Slowing of speech rate 16 36 27 2 28.2***
Stereotyped movements 12 30 19 1 29.6***
SYMPTOMS
Appetite loss 71 83 82 57 3.4***
Weight loss 59 78 69 45 3.1***
Slowed thinking 75 85 84 66 2.6***
Indecisive 86 82 93 83 1.6
Unpleasant thoughts 85 94 82 84 1.2
Slowed physically 82 83 91 75 2.6***
Suicidal thoughts 70 71 62 76 0.6*
Loss of interest
in pleasurable activities 94 96 96 88 2.1
Anticipatory anhedonia 87 93 94 80 3.8***
Consummatory anhedonia 84 92 95 72 5.8***
Incapacity to be cheered by
(a) pleasant event 81 90 92 70 4.7***
(b) social support 85 96 94 76 5.9***
Mood worse in morning 51 33 65 48 1.3
Energy worse in morning 48 42 60 43 1.5*
Terminal insomnia 52 64 59 42 2.1***
Non-variable mood 78 67 82 79 0.9
Constipation 46 72 55 30 3.5***
Pathological guilt 38 69 51 18 6.0***
f Clinical diagnosis of depression: PD = psychotic; ED = endogenous;
ND/RD = neurotic/reactive.
t Odds ratio for P D / E D versus ND/RD.
*** P < 0.001; ** P < 0.01; *18 P <:0.05
102 G. Parker and D. Hadzi-Pavlovic

Table 6.5. CORE II study of signs: post-rotation (Promax) factor


pattern (loadings) for the three-factor solution

Factor
Sign I II III

Slowed speed of movement 1.00 -0.04 -0.13


Slowing of speech rate 0.76 -0.05 0.02
Delay in motor activity 0.74 0.05 0.06
Body immobility 0.74 -0.11 0.20
Delay in responding verbally 0.52 0.07 0.32
Facial immobility 0.51 -0.02 0.33
Postural slumping 0.49 0.17 0.13
Facial agitation -0.01 0.86 -0.16
Motor agitation -0.08 0.78 0.05
Facial apprehension 0.11 0.75 -0.04
Stereotyped movements -0.12 0.62 0.16
Verbal stereotypy 0.10 0.60 0.13
Impaired spontaneity of talk 0.21 -0.14 0.79
Shortened verbal responses 0.19 -0.09 0.77
Non-interactiveness 0.02 0.12 0.75
Poverty of associations 0.18 0.08 0.74
Inattentiveness 0.10 0.31 0.43
Non-reactivity 0.34 0.30 0.39
Loadings greater than 0.35 are in bold.

the first factor was a general unipolar factor; the second was a bipolar
factor contrasting retardation and agitation items; and the third was
somewhat unclear.
Analysis of the three-factor solution suggested a potentially more clin-
ically useful delineation of psychomotor disturbance than the two-factor
solution, which merely defined retardation and agitation dimensions.
Factor loadings (see Table 6.5) suggested that the first and second fac-
tors clearly defined retardation and agitation dimensions respectively,
and, as their inter-factor correlation was 0.19, relative independence
of those domains was suggested. The third factor was labelled "non-
interactiveness" because the non-interactive item loaded highly, as did
three items reflecting verbal withdrawal (i.e., reduced length of verbal
responses, impaired spontaneity of talk and poverty of associations). As
the third factor inter-correlated 0.65 with factor I and 0.45 with factor
III, a non-interactiveness (or what we now view as a cognitive process-
ing) dimension appeared to underlie both the refined motoric retardation
and agitation dimensions.
Development and Structure of the CORE System 103
The status of this third factor (meaningful or artifact) appeared im-
portant to establish. Any such artifact could be a result of either the
initial third factor being needlessly retained and rotated or loadings on
the third factor fluctuating markedly. To examine the stability of this
third factor we formed 1000 bootstrap samples (random samples with
replacement) from our data, and carried out a factor analysis on each
of the samples. This procedure is detailed in Parker et al. (1993), with
results suggesting that while the third factor was small, it was both real
and stable.

Summary. While we had included only items with a priori weightings


(in our view) to agitation and retardation, our three-factor solution sug-
gested a trunk and branch model for construing psychomotor disturbance,
with a central non-interactiveness trunk (suggestive of a cognitive pro-
cessing domain) underpinning two arborising motoric domains of agita-
tion and retardation.

Frequency of Symptoms
By contrast with the signs (which, as noted, were affirmed on average
in only one in three of the depressed subjects), some symptoms (see
Table 6.4) appeared ubiquitous with, for instance, loss of interest in
pleasurable activities being affirmed by 94%, while indecisive, slowed
physically as well as several non-reactivity and anhedonia items were
affirmed by more than 80% - resulting in a high mean item prevalence
of 71% (compared to 31% for the signs). Again, as detailed in Ta-
ble 6.4, the overall chance of the PD/EDs being more likely than the
ND/RDs to affirm items was distinctly lower (median OR of 2.6) than
for the signs (median OR of 14.0), with non-significant ORs for a number
of individual items (i.e., indecisive, unpleasant thoughts, loss of inter-
est in pleasurable activities, mood worse in morning and non-variable
mood). While we failed to find evidence of specificity of symptoms to
the melancholic depressive disorders (i.e., clinically diagnosed PDs and
EDs) in such screening analyses, we nevertheless retained all symptoms
for examination in the multivariate analyses.

Summary. Overall, prevalence data demonstrated that the 18 symptoms


showed less suggested specificity to melancholia (at least as clinically
defined) than the set of 18 signs. A number of symptoms held historically
104 G. Parker and D. Hadzi-Pavlovic

to be weighted to a melancholic or endogenous type failed to show any


evidence of specificity.

Latent Class Analyses in the CORE II Study


While it is possible to use LCAs to reduce item sets (as occurred in the
CORE I studies), we chose not to do so here. As our aim at this stage
was to obtain a best estimate of the underlying classes, we judged that
all items should be included.

The Signs. We tested both two-class and three-class models, with items
scored (a) using all four response categories (0-3), and (6) by contrasting
the presence (0) or absence (1-3) of a sign. As both scoring methods
produced virtually identical solutions (all but 10 or 2.4% of subjects
being allocated to the same class), we used the raw ratings. As the
three-class solution was unstable, we report only the two-class solution.
Table 6.6 reports item probabilities as well as latent class proportions
(i.e., the proportions of the sample falling into each of the classes). As-
suming that Class A represents melancholia, the item probability is the
probability of any subject with melancholia rating positive on that item.
As noted earlier, if an item is specific to melancholia we would expect
a probability similar to its prevalence in the identified melancholic class
(which might be low for a rare feature) and a zero probability in the
non-melancholic class.
A large difference between an item's probability in Class A and Class
B points to the degree of specificity of that item. Large differential prob-
abilities of this kind are evident for the retardation scale items (e.g., 0.68
for body immobility) and for the non-interactiveness scale items (e.g.,
0.79 for poverty of associations). By contrast, the agitation items had
rather low probabilities in the melancholic class (e.g., 0.36 for facial agi-
tation, 0.44 for motor agitation) and with lower differential probabilities
(e.g., 0.18 and 0.28 for the two items noted). Low probability loadings
for agitation items are not necessarily a problem if agitation, by com-
parison with retardation, occurs only (as we believe to be likely) in a
minority of those with melancholia. Low differential probabilities for
the agitation items, however, is more of a concern, particularly if as-
sociated with a significant loading in the non-melancholic class (as was
evident for facial apprehension, facial agitation and motor agitation, but
not for verbal stereotypy and stereotyped movements). Such a finding
Development and Structure of the CORE System 105

Table 6.6. CORE II study: latent class item probabilities for the 18
signs and 18 symptoms, items scored 0-3

Class
Variable At BJ
SIGNS
Non-interactiveness 0.77 0.10
Facial immobility 0.88 0.22
Postural slumping 0.70 0.18
Non-reactivity 0.95 0.34
Facial apprehension 0.72 0.48
Delay in responding verbally 0.61 0.02
Shortened verbal responses 0.71 0.03
Inattentiveness 0.48 0.03
Facial agitation 0.36 0.18
Body immobility 0.74 0.06
Motor agitation 0.44 0.16
Poverty of associations 0.85 0.06
Slowed speed of movements 0.60 0.09
Verbal stereotypy 0.46 0.06
Delay in motor activity 0.43 0.01
Impaired spontaneity of talk 0.77 0.05
Slowing of speech rate 0.37 0.00
Stereotyped movements 0.24 0.02
Latent class proportions 40.4% 59.6%

SYMPTOMS
Appetite loss 0.83 0.53
Weight loss 0.74 0.42
Slowed thinking 0.86 0.66
Indecisive 0.93 0.83
Unpleasant thoughts 0.89 0.85
Slowed physically 0.93 0.70
Suicidal thoughts 0.30 0.67
Loss of interest in pleasurable activities 1.00 0.90
Anticipatory anhedonia 0.98 0.76
Consummatory anhedonia 1.00 0.69
Incapacity to be cheered by
(a) pleasant events 0.92 0.69
(b) social support 0.99 0.74
Mood worse in morning 0.50 0.47
Energy worse in morning 0.46 0.46
Terminal insomnia 0.65 0.35
Non-variable mood 0.71 0.78
Constipated 0.62 0.30
Pathological guilt 0.52 0.27
Latent class proportions 55.4% 44.6%
t Putative melancholic class. % Putativenon-melancholic class.
106 G. Parker and D. Hadzi-Pavlovic

could reflect (a) "agitation" not being specific to a melancholic class, or


(b) difficulties experienced by the raters in distinguishing melancholia-
specific agitation from non-specific motoric movements induced, say, by
anxiety or tension. The latter may be observed in many normal indi-
viduals under stress (hand-wringing being a good example) and thus is
not specific to melancholia. While both possibilities must be respected,
we favoured the latter (i.e., rater difficulties and errors in distinguish-
ing true psychomotor agitation) and made several important decisions.
Firstly, we elected to retain the agitation items in our final set of CORE
signs. Secondly, we modified some of our descriptors of agitation items
and interview recommendations, as detailed in the discussion.
Table 6.6 data indicate that the sign-based LCA allocated 40% of the
sample to a putative "melancholia" class.

The Symptoms. We attempted to fit two- and three-class models, but


only the former was interpretable. Table 6.6 reports item probabilities
and latent class proportions when items were scored 0-3. In contrast
to the LCA for signs, the LCA here demonstrated minimal specificity,
with only five items (appetite and weight loss, consummatory anhedo-
nia, terminal insomnia and constipation) having differential probabilities
greater than 0.30 in the melancholic class (compared to 14 of the signs),
and with most of the symptoms having fairly equivalent item probabili-
ties in each class (e.g., non-variable mood has loadings of 0.71 and 0.78)
and thus arguing against their specificity.
The symptom-based LCA allocated (55%) of the sample to a putative
melancholia class in contrast to 40% allocated by signs.

Summary. Both for the symptoms and for the signs, two-class solu-
tions (presumably generating melancholic and non-melancholic classes)
were stable and interpretable. Two requirements for clinical features to
be viewed as having specificity to a melancholic class - large differential
probabilities across melancholic and non-melancholic classes, as well as
minimal probabilities in the non-melancholic class - were met by all
signs bar several agitation signs, but were met poorly by all 18 symp-
toms. Such a limitation to some of the agitation items was viewed as
due to invalid ratings (and requiring more precise instructions), while
limitations across the set of symptoms argued for their limited capacity
(for whatever reason) to delineate a melancholic class.
Development and Structure of the CORE System 107
Concurrent Validity of the Derived Latent Classes. To check on
the credibility of the derived latent classes, we cross-tabulated alloca-
tions against (i) clinical diagnoses and (ii) for the MDU sub-sample,
DSM-III-R diagnoses (major depression with melancholia/psychotic fea-
tures vs. residual depressive diagnoses).
Using the LCA solution from the training sample, the success of class
assignment across systems was examined in both the training and vali-
dation samples (see p. 6 for definition). In relation to clinical diagnoses,
most subjects allocated to the sign-based melancholic class received a
clinical diagnosis of PD or ED (88% and 92% in the training and vali-
dation samples), while most allocated to the sign-based non-melancholic
class had a clinical diagnosis of ND or RD (77% and 76% in the two sam-
ples). The overall levels of agreement were similar in both the training
(ft = 0.63) and validation (K, = 0.67) samples, whereas equivalent analy-
ses for the symptom-based classes revealed lower levels of agreement (KS
of 0.40 and 0.48).
Next, in comparison with DSM-III-R allocations, the sign-based classes
showed low agreement (K,S of 0.33 and 0.42), with the DSM-III-R system
diagnosing a higher proportion of the sample than the sign-based LCA
to a melancholic class (57% vs. 34%). Equivalent analyses suggested
slightly higher (but modest) levels of agreement (KS of 0.39 and 0.53)
between the symptom-based classes and relevant DSM-III-R allocations.

Summary. The sign-based LCA allocations to a melancholic class cor-


responded strongly with a clinical diagnosis of psychotic or non-psychotic
melancholia, but modestly with DSM-III-R diagnoses. By contrast, the
symptom-based LCA correlated distinctly less well with clinical diagnoses
and modestly with DSM-III-R assignments. Further validation analyses
will be reported in Chapter IS.

A Possible Distortion by Psychotic Depression. It was possible


that the presence of the psychotic depressives (PDs) in our sample might
have distorted the analyses, and even that we might have only distin-
guished psychotic depression (rather then melancholia) from a residual
class. This possibility could be rejected if (a) the LCA probabilities were
unmodified when the analyses were undertaken with the PDs removed
(though these probabilities are not insensitive to prevalence) and (b) the
actual prevalences altered depending on the percentage of class members
removed.
Accordingly, the 43 PDs were left out of the training sample (now
108 G. Parker and D. Hadzi-Pavlovic

n = 207) and we estimated what the new prevalences were then likely
to be. As the melancholies represented 101/250 (40.4%) of the training
sample, the removal of 36 PDs (7 psychotics were classified as non-
melancholic) should have generated a new prevalence of 31.4% (65/207).
The LCAs for melancholic versus non-melancholic were re-run, with the
item probabilities proving to be similar to the original probabilities (as
reported in Table 6.3) and with the prevalences (67/207, or 32.4%) being
almost exactly as predicted.

Conclusion. We argue, from rather clear-cut analyses, that we had not


merely distinguished psychotic depression from a residual class in our
LCA of signs.

Deriving a CORE Score Index from CORE II


We derived a total CORE score by summing ratings of the 18 sign items.
Examined against clinical diagnoses, the combined PD/EDs returned
higher CORE scores (14.9 vs. 2.6, F = 434.0, P < 0.001) than the com-
bined ND/RDs, with the PDs in turn returning higher scores than the
EDs (19.2 vs. 12.6, F = 51.7, P < 0.001). Contrast analyses estab-
lished that those assigned by DSM-III-R as having major depression with
melancholia or psychotic features returned higher CORE scores (10.6 vs.
3.3; F = 41.9, P < 0.001) than residual categories. Finally, Newcastle-
defined endogenous depressives also had significantly higher CORE scores
(14.7 vs. 4.0; F = 99.7, P < 0.001). The distinct differences in CORE
scores for those with or without melancholia (however defined) is worthy
of emphasis.
While there was no sex effect, there was an age effect, with older
subjects scoring higher on total CORE scores (F = 8.03, P < 0.005).

Conclusions. However defined across three criteria sets, those with


melancholia (psychotic or not) returned substantive scores, with the non-
melancholic subjects scoring minimally on the CORE system. The age
effect is compatible with the chance of melancholia increasing with older
age.

Deriving a High CORE Category


Using the LCA melancholia allocation as our criterion measure, we cal-
culated a cut-off CORE score for assigning subjects as having "melancho-
Development and Structure of the CORE System 109
lia," using (as in the CORE I study) the receiver operating characteristics
(ROC) associated with varying cut-offs (or decisions). The optimal cut-
off in CORE score was established as 8 or more, and agreement with the
classes denned previously in the LCA was high in both the training and
validation samples, with kappas of 0.92 and 0.89 respectively.
We established only a modest level of agreement (n = 0.36) between
melancholia defined by DSM-III-R and those scoring 8 or more on the
CORE measure. Specifically, of 183 available cases, both systems rated 53
(29%) as melancholic and 69 (38%) as non-melancholic, with a further 52
(28%) rated as melancholic by DSM-III-R but non-melancholic by CORE,
and 9 (5%) the converse. Presuming only one system to be valid, then
either the CORE system was under-diagnosing true melancholia (with a
rate of 34% in this sub-sample) or the DSM-III-R criteria set was too
inclusive (with a sample rate of 57%). We pursue the validity of each
system in several later chapters but note here an important analysis
using the dexamethasone suppression test (DST).
If, as claimed, the DST is a biological index of melancholia, the com-
parative utility of several measures of melancholia can be tested. We
therefore studied a sub-sample of those with DSM-III-R melancholia and
compared DST non-suppression rates for sub-samples of (a) high CORE
and (6) low CORE subjects. Examining the 8 A.M. data, 15/33 (45%)
high CORE and 1/11 (9%) low CORE subjects were non-suppressors, while
the combined 0800 h and/or 1600 h data showed a similar non-significant
trend - 21/32 (66%) and 6/11 (55%) respectively. Such findings indi-
cated that, within the subjects who already met DSM-m-R criteria for
melancholia, those with significant psychomotor disturbance were much
more likely to be DST non-suppressors. In Chapter 8 we report an
extension of this DST study demonstrating the comparative superior-
ity of the CORE system melancholic allocations (compared to DSM-III-R
assignment) when examined using DST non-suppression rates.

Conclusions. Despite any imprecision imposed by forcing a cut-off, a


CORE score of 8 or more was in strong agreement with the LCA allo-
cation to a melancholic class. However, a CORE score of 8 or more
was poorly linked with DSM-III-R melancholia assignment. We use DST
results to argue the comparative superiority of CORE assignment in gen-
erating a more homogeneously refined group of depressed patients (pre-
sumed melancholia).
110 G. Parker and D. Hadzi-Pavlovic
Deriving a SYMPTOM Score. As for the signs, we summed symp-
tom ratings to generate a total SYMPTOM score, and established that to-
tal SYMPTOM scores were positively associated with patient-rated Zung
scores (r = 0.45) and older age (r = 0.18) but independent of sex. The
ROC strategy was similarly applied to symptom scores, with a cut-off
of 26 or more determined as distinguishing likely melancholia from non-
melancholia. Agreement with the symptom latent classes was examined
in the training and validation samples, with the kappas (0.81 and 0.75)
lower than equivalent analyses for the CORE signs.

Mixture Analyses of CORE and SYMPTOM Scores. We ex-


amined the evidence for separate sub-populations in CORE scores and
SYMPTOM scores in the total sample of 413 subjects using mixture anal-
ysis (Agha and Ibrahim 1984). For SYMPTOM scores, the test for 1
versus 2 populations was significant (x2 = 14.7, df 4, P < 0.005) while
the test for 2 versus 3 populations was not significant (x2 = 5.3, df 4,
P = 0.25). Such results suggested that SYMPTOM scores were a mixture
coming from two underlying populations.
For CORE scores the test for 1 versus 2 populations was significant
(x2 = 257.8, df 4, P < 0.001), as was the test for 2 versus 3 popula-
tions (x2 = 55.3, df 4, P < 0.001). These results suggest that CORE
scores were a mixture of three underlying populations. One population
appeared (by reference against clinical diagnostic data) to reflect non-
melancholic subjects. The remaining two populations comprised either
psychotic versus non-psychotic melancholia or retarded versus agitated
melancholia, but were not further clarified at this time. Analyses re-
ported in Chapter 12 favour the first explanation.

Does Psychomotor Disturbance Alone Define Melancholia?


Having focussed on one observable behavioural construct and system-
atized its components, we next examined (Parker et al. 1995a) the com-
parative utility of psychomotor disturbance (PMD) against historically
suggested endogeneity symptoms as predictors of melancholia. Thus, we
examined a more refined hypothesis: in depressed subjects, PMD is both
necessary and sufficient for clinical definition of melancholia.
Any testing of the centrality of PMD to melancholia is nevertheless
limited by a number of realities - mainly that there is no gold standard
measure of melancholia - so that the extent to which the principal mea-
Development and Structure of the CORE System 111
sures used in this study component (DSM-III-R and Newcastle criteria)
validly circumscribe melancholia must be conceded at the outset.

Introduction and Characteristics of the Somewhat Refined Sam-


ple. In this section of studies, we excluded 77 subjects who were rated
at baseline as neither at nor close to the nadir of the episode (and for
whom cross-sectionally assessed sign-based data may not have validly
reflected their status) and 3 who did not meet DSM-III-R major depres-
sive episode (MDE) criteria, resulting in a final sample of 327 patients
with a base diagnosis of MDE.
In addition to the 46 subjects who were assigned by DSM criteria
to a psychotic depression class, we had variable groupings of melan-
cholic/endogenous and comparison groups. DSM criteria assigned more
subjects (148) to a non-psychotic melancholic class than our equivalent
clinical diagnosis of endogenous depression (84), with both exceeding
the 62 assigned as endogenous by the Newcastle Scale, and with such
differential frequencies (45%, 26% and 19%) questioning the validity of
each of these three estimates of melancholia. In our analyses we priori-
tized the two formal diagnostic systems (DSM-III-R and Newcastle) over
our (conceivably biased) clinical diagnoses.

Examining the Necessary and Sufficient Hypothesis. If PMD is


necessary to the definition of melancholia, then all melancholies should
return the pre-established CORE score of 8 or more. This appeared
reasonably supported against Newcastle (53/62, or 85%) and against
clinical diagnoses (64/84, or 76%) but not against DSM-III-R (75/148,
or 51%) diagnoses. If PMD is sufficient, then all depressives with a
CORE score of 8 or more should be assigned to a melancholic diagnostic
class (while those without PMD might or might not be). The sufficiency
criterion was largely met against DSM-III-R, with 75/88 (85%) meeting
the CORE cut-off being assigned to DSM-III-R melancholia, but less well
for Newcastle (53/88, or 60%) and clinical (64/88, or 73%).

Conclusions. The Newcastle and clinical reference assignments pro-


duced rather similar results, largely meeting the necessary component
but only moderately satisfying the sufficient criterion. Against DSM-III-
R assignment, the opposite pattern was demonstrated. Results did not
support the claim that PMD is both necessary and sufficient to defini-
tion of melancholia, and suggested that other contributing features (i.e.,
endogeneity symptoms) should be assessed as well.
112 G. Parker and D. Hadzi-Pavlovic

Refining Potential Endogeneity Symptoms. As our earlier analy-


ses had suggested that a number of our 18 symptoms showed poor speci-
ficity to a melancholic class, we now sought to refine our list of likely
endogeneity features to those most likely to make some independent
and/or additional contribution (cf. PMD) to the definition of melancho-
lia. Our logic was that

(i) any endogeneity feature should not only show some specificity to
an endogenous/melancholic class, but that
(ii) such specificity should be sustained across both the DSM-III-R and
Newcastle diagnostic systems (i.e., over-ride any imperfections of
each system in circumscribing melancholia), and that
(iii) the frequency of any such item should reflect the prevalence of
melancholia or melancholic sub-type (e.g., the frequency of delu-
sions should not exceed the frequency of any psychotic melancholia
sub-type).
We therefore excluded a number of non-discriminatory symptoms, ei-
ther because they were commonly reported by depressives in all groups
(i.e., slowed thinking, indecisive, trouble concentrating, unpleasant
thoughts, slowed physically) or because they had a similar frequency
in each group (suicidal features). A correlation matrix of the 12 re-
maining provisional endogeneity symptoms indicated a number of strong
associations. Thus, appetite and weight loss were significantly intercor-
related (r = 0.61), as were morning worsening in mood and energy levels
(r = 0.68) and anticipatory and consummatory anhedonia (r = 0.58),
allowing some amalgamation of those 6 individual variables to 3 com-
posite ones. We thus examined a list of 9 refined endogeneity symptoms.

(1) appetite and/or weight loss; (2) non-reactive mood;


(3) anticipatory and consummatory (4) terminal insomnia;
anhedonia; (6) loss of interest;
(5) diurnal variation in mood and (8) sinful/guilty; and
energy; (9) sense of deserving
(7) constipation; punishment.

Summary. We refined our list to nine consolidated endogeneity symp-


toms on the basis of their showing some specificity to the melancholic
class and being independent of each other to some extent.
Development and Structure of the CORE System 113
Are Endogeneity Symptoms Necessary to the Diagnosis of Mel-
ancholia? If PMD (as defined and quantified) is a reasonably efficient
predictor of melancholia status, this still leaves the question about which
other clinical features improve assignment to the melancholic class. As
most diagnostic systems focus on endogeneity symptoms in such deci-
sions, we examined the extent to which our nine refined and composite
endogeneity symptoms improved allocations to melancholic classes above
and beyond PMD alone.
In separate multivariate analyses (discriminant function and logistic
regression) we examined (Parker et al. 1995a) the predictive capacity
of the following three variables to discriminate melancholic and non-
melancholic allocations as generated by the three diagnostic systems:
(i) CORE score alone
(ii) total score on the nine endogeneity symptoms (SYMPTOM) alone
(iii) both CORE and SYMPTOM scores.
As quantification was very similar across both multivariate strategies,
we report only the logistic regression results. Examining kappa coeffi-
cients of agreement, analyses indicated that:
(i) DSM-III-R assignment was predicted equally by CORE score alone
and by SYMPTOM score alone (K — 0.43 for both).
(ii) Newcastle assignment was predicted far more strongly by CORE than
by SYMPTOM score (K = 0.63 and K = 0.14).
(iii) Clinical diagnostic assignment showed a trend similar to the New-
castle results (K = 0.56 and K = 0.24).
Importantly, when CORE and SYMPTOM scores were entered together,
the combined predictors added little to the prediction generated by the
CORE score alone (i.e., the increments in kappas being 0.06 for DSM,
0.02 for Newcastle and 0.01 for Clinical). Conversely, when CORE scores
were entered after SYMPTOM scores, the kappa increments were again
slight for DSM (at 0.06) but substantive for Newcastle (0.51) and Clin-
ical (0.33), indicating that (apart from the DSM system) melancholia
definition was considerably improved by the additional CORE predictor.
We pursued the sufficient component of the hypothesis further by ex-
amining any additional predictive power effected by adding each of the
nine refined endogeneity symptoms as individual categorical (i.e., present
vs. absent) predictors - in comparison to the preceding approach of ex-
amining their combined influence as a summed dimensional SYMPTOM
score. Only four symptoms (appetite/weight loss; diurnal mood/energy
114 G. Parker and D. Hadzi-Pavlovic
change; terminal insomnia; anhedonia) made a significant improvement
(above PMD alone) in predicting melancholia - with the first three con-
tributing to DSM-III-R assignment (not unexpected, as these symptoms
are included in the DSM-III-R diagnostic criteria for melancholia), and
with diurnal mood/energy change and anhedonia contributing to both
Newcastle and Clinical assignments.

Conclusions. While the numbers assigned as having non-psychotic mel-


ancholia by the three varying systems varied considerably (62 to 148
respectively), the analyses suggested that CORE scores were largely suffi-
cient in and of themselves to predict melancholia allocations.

Shared Variance Examined


The last conclusion - that CORE scores (as a measure of behaviourally
judged PMD) alone define melancholia - is a little forced, as there is
clearly an overlap between what is measured by the CORE system and a
number of the endogeneity symptoms. We sought then to quantify such
shared variance in a number of analyses.

(i) Examining summed endogeneity SYMPTOM scores first, we estab-


lished a moderate overall correlation (r = 0.46, P < 0.001) with
total CORE scores.
(ii) Associations between CORE and individual endogeneity symptoms
were then examined in three ways:
(a) Undertaking correlational analyses, with indicative associations
being in the 0.25-0.50 range.
(b) Estimating the probability of a patient with a high CORE score
(i.e., 8 or more) reporting a symptom as more likely to be present
with any over-representation quantified by ORs as follows: ap-
petite and/or weight loss (3.7); loss of interest (3.7); anhedonia
(3.3); terminal insomnia (2.7); non-reactive mood (2.6); consti-
pation (2.6); sinful/guilty (2.0); diurnal mood variation (1.5).
(c) Undertaking logistic regression analyses, where the prob-
ability of a symptom being present was modelled using increa-
sing levels of grouped CORE scores, and which demonstrated (see
Figure 6.1) some distinctly differing patterns. Both diurnal vari-
ation in mood and/or energy, as well as acknowledging feeling
deserving of punishment, showed only a slight progressive in-
Development and Structure of the CORE System 115
crease as the CORE score increased. The chance of reporting
each of the remaining symptoms increased with an increase in
CORE scores, with the presence of several being almost inevitable
at some level of CORE-defined PMD. Some symptoms (e.g., non-
reactive mood) were infrequent with low CORE scores, while oth-
ers (e.g., appetite and/or weight loss in particular) were present
in a significant percentage of subjects without any PMD - al-
though their prevalence increased as CORE scores increased.

Conclusions. The overall correlation between summed CORE and re-


fined endogeneity scores (r = 0.46,) showed the shared variance to be lit-
tle more than 20%. Some endogeneity symptoms (e.g., appetite and/or
weight loss; loss of interest; anhedonia) were more closely linked with
CORE scores than others. However, logistic regression analyses demon-
strated that the inter-relationships between the two domains (of CORE-
defined psychomotor disturbance and endogeneity symptoms) were not
straightforward, being clearly influenced both by each individual symp-
tom and by the degree of psychomotor disturbance.

Comparison of Diagnostic Indices and Systems


In our principal report of the CORE II study (Parker et al. 1994) we
examined the extent to which 7 varying definitions of melancholia dis-
criminated between a number of study variables, and supported the
validity of each definition. Such an examination was limited to the 185
MDU patients, where our baseline assessment considered a wider range
of variables, and from which independent (DST and neurocognitive re-
action time) data had been sampled. The derived melancholic classes
comprised:

(i) those assigned by the sign-based LCA (SIGN LCA)


(ii) those assigned by the symptom-based LCA (SYMPTOM LCA)
(iii) those categorised as HIGH CORE subjects (i.e., receiving a CORE
score of 8 or more)
(iv) those categorised as HIGH SYMPTOM subjects (i.e., receiving a SYMP-
TOM score of 26 or more)
(v) those meeting DSM-III-R criteria for major depression with melan-
cholia or psychotic features (DSM-III-R)
116 G. Parker and D. Hadzi-Pavlovic

20 25

Total score on CORE

Fig. 6.1. Probabilities of Symptoms Being Affirmed Modelled Using Logistic


Regression. Adapted from Parker et al. (1995a).
Development and Structure of the CORE System 117
(vi) those receiving a clinical diagnosis of PD or ED (CLINICAL)
(vii) those receiving a score of 6 or more on the Newcastle system (NEW-
CASTLE)

As we report a similar comparative analysis in Chapter 13 (where


we also examine the comparative utility of an empirically derived algo-
rithm), we merely summarise several key findings here. If merely the
number of significant differences is considered, then:

(i) the HIGH SYMPTOM allocation was the least successful (with only 7
variables significant).
(ii) of the remaining systems, assignment by the SYMPTOM LCA was
the least discriminating (with significant differences for only 9 vari-
ables).
(iii) the remaining systems (DSM-III-R, CLINICAL, the SIGN LCA, the
CORE score and NEWCASTLE) performed rather similarly, discrimi-
nating on 12-14 variables. By each, assigned melancholies:
(a) were older;
(b) were older at initial episode;
(c) had a briefer current episode;
(d) were more severely depressed on the Hamilton scale (but not
on the self-rated Zung scale);
(e) were more likely to acknowledge a clear remission for a previous
episode (and a distinct improvement following any ECT course);
(f) were more likely to have delusions and hallucinations; and
(g) were more likely to have demonstrated slowed neurocognitive
reaction time patterns.
The CLINICAL, DSM-III-R and SIGN LCA-assigned melancholies had
significantly higher DST non-suppression rates at 0800 h and/or
1600 h, while, for the 0800 h data, only the sign-based assignments
(SIGN LCA and HIGH CORE) showed significantly differential non-
suppression rates. On each of three neurocognitive reaction time
measures, both the HIGH CORE and SIGN LCA-defined melancholic
subjects showed the greatest differentiation of mean scores.
(iv) The analyses for the sign-based systems were repeated with
the 20 clinically diagnosed PDs deleted, and there was a minimal
impact on results. Thus, we concluded that these two systems dis-
118 G. Parker and D. Hadzi-Pavlovic
criminated melancholia rather than defining, or being significantly
influenced by, clinical features of psychotic depression.

Conclusions. If a melancholia index is valid, it should differentiate de-


pressed subjects on a number of the clinical givens of melancholia. To
be useful, any such index must be as discriminating or superior to other
indices or diagnostic criteria sets. The current provisional analyses (ex-
tended in Chapter 13) indicated that high CORE score assignment was
as successful as the most successful comparative systems, and certainly
more successful than the two systems based on endogeneity symptoms
alone.

Discussion of the CORE II Studies


We seek to revive interest in the psychomotor disturbances noted in
melancholia historically over time, and to establish whether an obser-
vationally based measure of psychomotor disturbance may distinguish
melancholic from non-melancholic depression with some precision. Our
CORE I studies provided evidence to suggest superiority of such an
approach over symptom-based systems. Those findings were replicated
and extended in the CORE II studies, which led to development of the
final CORE measure.

Notes on the Development of the Final CORE Measure. We


preserved a number of items generated in the CORE I study, added
several others (especially agitation items) and then deleted items on the
basis of empiricalfindings- resulting in a refined set of 18 items assessing
observer-rated signs of psychomotor disturbance.
A principal components analysis of those 18 items suggested a trunk
and branch analogy for (a) construing psychomotor disturbance and for
(b) deriving sub-scales (see Figure 6.2). The six truncal items contribute
to a non-interactiveness scale, which, on the basis of item inspection,
appears to reflect a neurocognitive processing dimension or the psychic
component of psychomotor disturbance, and which, when severe, may
generate a clinical diagnosis of pseudo-dementia in depressed patients.
Importantly, correlational analyses indicated that disturbances in neu-
rocognitive processing underpin those with either significant agitation
or significant retardation.
The remaining CORE items define motor aspects of psychomotor dis-
turbance, with some specificity either to agitation or to retardation.
Development and Structure of the CORE System 119

Facial agitai •wed speed of movement


Motor agitai wing of speech
Facial apprehen: in motor activity
Stereotyped movi lily immobility
Verbal stereotyj iay in responding verbally
ial immobility
ural slumping

Non ivity
Inatt ness
Poverty ociations
Non-int eness
Shortened I responses
Impaired neity of talk

NON-INTERACTIVENESS ITEMS

Fig. 6.2. Tree and Trunk Model for Construing Psychomotor Disturbance and
the CORE'S Sub-scales.

While raw scores on the generated agitation and retardation scales were
non-significantly associated with each other (though each was signifi-
cantly associated with scores on the non-interactiveness sub-scale), we
need to clarify the implication of these relationships in applied research
(especially pursuing aetiology and treatment response).
Both the inter-factor correlations from the factor analysis and the
correlational analysis of the subsequent agitation and retardation scale
scores indicated that there was no relationship between these scales.
Such apparent independence was not forced by us, as the oblique rota-
tion of factors allowed for them to be correlated. This does not mean that
agitation necessarily occurs independently of retardation in the clinical
context. We suggest, from clinical observation, that those with so-called
retarded melancholia show only retardation, while those with agitated
120 G. Parker and D. Hadzi-Pavlovic
melancholia have a base of retardation upon which epochs of agitation
are superimposed (a non-independent relationship). Such realities may
thus result in higher CORE scores being generated in those with an agi-
tated melancholia than in those with retarded melancholia.
While thefiveitems contributing to the agitation scale formed a coher-
ent component in the three-factor factor analysis solution, their capacity
to demonstrate specificity to the melancholic class was less clear in the
subsequent LCA. Thus some of the agitation items (a) had differential
item probabilities across Classes A and B that were not particularly im-
pressive (see Table 6.6), and (b) had significant probability loadings in
Class B, when non-existent or trivial loadings would be expected if they
had specificity to the melancholic class.
We suspect that such results emerged from our relative failure to de-
fine melancholia-specific psychomotor agitation at the commencement of
the study, and predictably problematic ratings as a consequence. Once
aware of the difficulty (after inspecting the LCA results), we set out to
redress the problem by providing more phenomenologically distinct and
rich definitions. We have had some success in our progressive modifica-
tion of the descriptors (see Appendix) and of our general instructions to
raters (see Chapter 14). Residual difficulties reflect our suspicion that
some expressions of agitation may occur in a range of circumstances -
not limited to melancholia (or even to depression). Thus, we cannot
distinguish phenomenologically between the hand-wringing of a patient
with melancholic depression and (say) that of a non-depressed medi-
cal student experiencing a distressing viva voce examination or that of
a Hollywood actor waiting for an Oscar award to be announced. We
have greater confidence that we can capture melancholia-loculated char-
acteristics in our item definitions and operational criteria as agitation
moves to the presence of stereotypies. Thus, when the individual is not
merely hand-wringing but tearing at the fingers, or plucking at parts of
the body, or expressing certain verbal stereotypies (e.g., "What is going
to become of me?"), we feel more confident in viewing such behaviours
as melancholia-based and able to be distinguished from tension-driven
mannerisms and other non-melancholia-related repetitive movements in
our item descriptors.
As phenomenological definition of the generic agitation items remains
problematic, we have addressed the problem in a somewhat indirect
manner. Clinical observation suggests that the agitation associated with
melancholia is largely impervious to reassurance (with the patient re-
maining either persistently agitated or having epochs of agitation across
Development and Structure of the CORE System 121
the interview), while agitation produced by the stress of the interview
settles after 10-15 minutes in a supportive environment. Therefore,
raters are now instructed (see Chapter 14 and Appendix) to undertake
their "normal clinical intake interview," during which they should at-
tempt to establish rapport and settle the patient, and that they should
allow a minimum of 20 minutes before rating. Additionally, as our clin-
ical observations suggest that melancholia-associated stereotyped move-
ments do not occur in the absence of motor agitation, our instructions
allow stereotyped movements to be rated only if the patient has been
rated as having motor agitation.
At the practical level, these several strategies result in those with
agitation generated by stress and tension returning CORE scores of 2-6
if the period of settling the patient has not met the objective of resolving
such agitation. As the suggested CORE cut-off score for melancholia is
8 or more, the chance of falsely allocating a subject with an anxious
non-melancholic depression to the melancholia class is then remote.

Use of the CORE Measure in Applied Studies. We suggest that


the CORE measure may be used in three principal ways:
(i) as a total CORE score, and thus a dimensional measure of psychomo-
tor disturbance;
(ii) as three-dimensional sub-scale scores (i.e., non-interactiveness, agi-
tation and retardation); and
(iii) to generate categories - either of likely melancholic and non-melan-
cholic depression or, as detailed in Chapter 12, of likely psychotic
depression.
In relation to (i), we have noted (Parker et al. 1994) that total CORE
scores are independent of sex, and that higher scores are returned by
those with psychotic depression (in particular) and those with non-
psychotic melancholia (however defined), while non-melancholic de-
pressed subjects return minimal scores.
In relation to (ii), we have established (Parker et al. 1993) that sub-
scale scores are also independent of sex and increased in those with
psychotic and non-psychotic melancholia, and, for those who have had
a previous episode, higher scores indicate a greater chance of reporting
improvement with either an antidepressant drug or ECT.
In relation to (iii), we focus here on the utility of a CORE score cut-off
of 8 or more (HIGH CORE) as an estimate of likely melancholia. Examin-
ing concurrent validity first, agreement between HIGH CORE assignment
122 G. Parker and D. Hadzi-Pavlovic
and assignment to equivalent melancholia diagnoses was established as
moderate against both clinical diagnosis and Newcastle allocation, but
poor in relation to the DSM-m-R melancholia criteria. Nevertheless,
comparative analyses (Parker et al. 1994, and considered in Chapter 13)
demonstrated that HIGH CORE assignment was as successful as many
other melancholia indices when examined against a set of validators.
Specifically, such depressed subjects were older and older at age of first
episode, were more likely than non-melancholies to have a bipolar illness
course, had a significantly higher rate of both delusions and hallucina-
tions during the episode and were more likely both to report a clear
remission and a distinct improvement with ECT if they were so treated
during a previous episode - features corresponding with the literature
on melancholia (Nelson and Charney 1981; Zimmerman et al. 1986b).
Additionally, the DST data provided laboratory study support for CORE
allocation being at least as discriminating as previously established dis-
criminating systems (Zimmerman et al. 1985).
Our comparative analyses offered strong support for the CORE system
as a measure of melancholia. However, as each of the systems provided
quite differing prevalence estimates of "melancholia" within the one sam-
ple, one measure at least has to have a significant error component; we
therefore examine the comparative utility of the CORE in subsequent
chapters.
At this stage, its comparative success - while akin to that for several
other measures - is impressive when it is remembered that the CORE
system is limited to the assessment of psychomotor disturbance only -
in effect, a cross-sectional "behavioural snapshot." By contrast, both
the DSM and Newcastle systems allow the clinician to include longitudi-
nal factors in the decision process. For instance, DSM-m-R criteria for
melancholia include, in addition to symptom criteria, two course of ill-
ness items (recovery from prior episode and history of good response to
somatic antidepressant treatment) and one pre-morbid feature (absence
of pre-morbid personality disorder).

Why Signs Might be a More Accurate Measure of


Melancholia than Symptom-based Approaches
We argue this issue from a number of pieces of evidence, all underpinned
by the reality that clinical definition of a depressive type requires demon-
stration of type-specific features.
Development and Structure of the CORE System 123
Comparative Prevalence Data of the Individual Endogeneity
Symptoms and Signs. As noted earlier, if a clinical feature has speci-
ficity, then its frequency should not exceed the prevalence of melancholia
in that particular sample. While there is no "gold standard" measure
of melancholia, we had three estimates of melancholia (including those
with psychotic features) - some 52% according to clinical diagnoses, 57%
according to DSM-III-R and 32% according to the Newcastle cut-off allo-
cation (the latter two being in relation to the MDU subjects only). To
the extent that these estimates have any validity, we might expect that
some 50% of our sample had melancholia. We would therefore not ex-
pect type-specific clinical features to exceed a 50% prevalence. Table 6.4
data demonstrated that only 3 of the 18 signs had a frequency above
50% (with the highest frequency item being non-reactivity at 61%). By
contrast, 15/18 symptoms exceeded a sample prevalence of 50% (with 12
of those exceeding 70%, and 8 exceeding an 80% prevalence). In general
terms then, most of the signs met this first requirement, while most of
the symptoms did not.

Demonstration of Specificity to Type. Clinical features which are


type-specific for melancholic type should show this in two ways - by
having a significantly differential prevalence or differential probability
across melancholic and non-melancholic types, and by being rare in the
latter. Assuming that our LCAs of symptoms and signs assigned those
with true melancholia to Class A and those with true non-melancholia to
Class B, then we can test for evidence of such specificity. Our assumption
was supported in that we demonstrated (Table 2, Parker et al. 1994) that
those allocated to Class A either by the symptom- or sign-LCA met a
number of the clinical givens of melancholia. They were older, were older
at first episode, had a more severe depression, and were more likely to
have delusions and hallucinations. While Table 6.6 data demonstrated
differential item probabilities of 0.30 or more to Class A for the majority
of the signs (14/18), only a minority of the symptoms (5/18) exceeded
that differential. If we impose an upper limit of 0.10 to the probabilities
as failure or relative failure to load in the non-melancholic class (Class
B), then only 6/18 signs exceeded that probability as opposed to all of
the symptoms. The mean probability for the 18 symptoms was 0.61 in
Class B, with 17/18 exceeding 0.30. In terms of raw prevalence and
item probabilities, the symptoms demonstrated unimpressive specificity,
generally being reported by most depressed patients.
Thus a number of the so-called endogeneity symptoms were virtually
124 G. Parker and D. Hadzi-Pavlovic
ubiquitous across the sample, suggesting (a) that many of them may
better define depression per se (and its severity) than any melancholic
depressive type, as well as (6) limitations to assessing such clinical fea-
tures by use of patients' self-reports. The former issue is difficult to
clarify, the latter somewhat easier by reference to the actual data. If we
assume that psychomotor disturbance is a clear marker of melancholia,
then those symptoms which appear to assess psychomotor disturbance
most directly (e.g., slowed thinking, indecisive and slowed physically)
failed to demonstrate clear differential probabilities or specificity. We
found that 82% of the sample rated themselves as slowed physically.
However, when speed of movement was rated as a sign (i.e., based on
behavioural observation by a clinician), only 32% were judged by the
clinician as being "slowed." While poor agreement does not necessarily
invalidate an observation - as patients and clinicians may use differing
reference points reflecting differing valid domains of relevance or con-
cern - it appears that only such behavioural ratings allow the potential
specificity of features such as psychomotor disturbance to be captured.
Such a finding (while rarely examined) is not unique to our sample. In
our re-analysis of Aubrey Lewis' data set (Parker and Hadzi-Pavlovic
1993) we noted a similar disparity in the LCA variables - with retarda-
tion judged observationally (generally by nursing staff in Lewis' study)
being far more discriminating across the two classes than self-reported
retardation (i.e., probabilities of 0.57 vs. 0.00 for observational, and 0.56
vs. 0.65 for self-reported retardation).
Whatever the explanation - that endogeneity symptoms lack speci-
ficity or that any self-report strategy is an invalid methodological ap-
proach - such a substantive affirmation rate for most of the endogeneity
symptoms by our non-endogenous/melancholic patients offers a major
challenge to symptom-based data as a strategy for the differential diag-
nosis of melancholia.

Evidence from Total Scores. As possible total symptom and sign


scores may range from 0 to 54, it is of interest that our ROC analy-
ses established quite different cut-off scores (i.e., 26 or more for symp-
toms; 8 or more for signs). As noted earlier, those diagnosed as having
melancholia by each of the three diagnostic systems returned substan-
tive CORE scores, while the non-melancholic subjects returned trivial
scores, suggesting a categorical model (subjects either had or did not
have psychomotor disturbance). By contrast, the high cut-off for symp-
tom scores indicated a more dimensional model - all depressives scored
Development and Structure of the CORE System 125
on the symptom scale to some degree, but those with melancholia tended
to score higher.

Evidence from Mixture Analyses. In establishing supportive evi-


dence of two sub-populations in SYMPTOM scores we found means of
16.8 in the putative non-melancholic and 30.1 in the putative melan-
cholic subjects, again suggesting that SYMPTOM scores lacked speci-
ficity (otherwise a trivial score would have been returned for the non-
melancholic subjects), and at best reflected a mix of severity and
depression sub-typing components.

The Necessary and Sufficient Hypothesis. We reported a num-


ber of analyses indicating that CORE scores were superior to symptom-
based scores in predicting endogeneous depression as defined by both the
CLINICAL and Newcastle diagnostic systems, although each predicted as-
signment to DSM-lii-R melancholia to a rather similar degree. In those
analyses we had refined our original list of 18 symptoms down to 9 symp-
toms which, in terms of prevalence, appeared to have greater specificity
to the melancholic class.
Such analyses suggested major doubts about the capacity of symptom-
based approaches to provide clinical definition of melancholia. While our
sign-based approach appears superior, its utility and validity require
close testing, as pursued in subsequent chapters.

Can a Sign-based System Define Melancholia?


The question is contrived, as we do not propose that CORE scores alone
should dictate diagnostic decisions, believing that other clinical features
in the first instance and, in the future, other criteria (e.g., aetiological
factors, biological markers, etc.) may be particularly useful to a diag-
nostic algorithm. Nevertheless, such a strict postulate is worth pursuit if
psychomotor disturbance is specific to the melancholic type of depression
and if it can be measured reliably and validly.
We address this issue in several ways. First, by essentially considering
alternative explanations. Thus, we consider whether CORE estimates of
melancholia may have been confounded by (a) psychotic depression, (b)
severity or (c) chronicity and treatment effects.

Confounding by Psychotic Depression. Two sets of analyses ad-


dressed this possibility: (i) whether the clinically diagnosed psychotic
126 G. Parker and D. Hadzi-Pavlovic
depressives (PDs) were included or excluded, our LCA-defined melan-
cholia class had similar item probabilities for the signs; (ii) when we re-
peated our validation analyses with the clinically diagnosed PDs deleted,
discrimination across most variables persisted. Such findings allow two
interpretations: first, that psychomotor disturbance is a feature common
to melancholic and psychotic depression but that it is more pronounced
in the latter; secondly, that psychotic depression and melancholia are
not separate types but differ principally in severity, an issue pursued in
Chapter 12. However, even if psychotic depression is a separate type of
depression, our analyses here indicate that any confounding by the in-
clusion of those with PD has not influenced our principal interpretative
analyses.

Confounding by Severity. Might we have merely defined a more se-


vere type of depression rather than a separate type of depression (qua
melancholia)? Varying definitions of severity might generate differing
conclusions. In this study, as in the CORE I studies, CORE-defmed
melancholies and non-melancholies did not differ on the Zung scale, a
self-report severity measure. By contrast, CORE-defmed melancholies
scored more highly on the Hamilton, although that scale confounds
severity and depressive type (i.e., higher scores will be driven both by
severity affecting non-specific items such as depressed mood, and by
items weighted to depressive sub-types, e.g., psychotic features, patho-
logical guilt).
Some theorists (e.g., Widlocher 1983) have suggested that retardation
is a primary behavioural disturbance in all affective disorders, building
to a dimensional view which equates the level of psychomotor change as
a measure of overall severity, quite independent of depressive type. If
true, or if we have confounded CORE estimates by a unitary dimension
of severity, we would expect that the non-melancholic depressives should
return reasonably substantive scores on the basis of severity, albeit not
as high as scores returned by those with melancholia. Instead, the mean
score for the latter was low, so that there was a sharp "trend break"
(or non-linear relationship) across the melancholic and non-melancholic
groups. Thus, while we accept that (generally and globally) melancholia
is a more "severe" depressive type, we argue that our findings favour
psychomotor change as specific to melancholia and not merely a non-
specific measure of depression severity.
Development and Structure of the CORE System 127
Confounding by Treatment Effects. Could we have confounded def-
inition of melancholia with treatment effects, when any pre-baseline use
of antipsychotic or even antidepressant medication might influence rat-
ings of psychomotor change? We have some indirect data on this point.
In Chapter 12 we note in the CORE I sample that the psychotic de-
pressed patients were equally likely (compared to the endogenous de-
pressed patients) to be prescribed antipsychotic medication and were
less likely to be receiving antidepressant medication at baseline assess-
ment, and concluded that higher CORE scores for the PDs were then
a reflection of differences between the underlying disorders rather than
medication effects on scores. Nevertheless, as noted, this is somewhat
indirect information. While the problem is not unique to a measure
such as the CORE, as a number of depressive and endogeneity symptoms
are clearly influenced by medication effects, any effects of psychotropic
medication on CORE scores do require separate study and quantification.

A "Necessary and Sufficient" Hypothesis. A second, more direct


way of addressing whether the CORE system could define melancholia
was to pursue the hypothesis that PMD is both necessary and sufficient
to the clinical definition of melancholia. While attractive for its parsi-
mony and potential to explicate neurobiological determinants, its pur-
suit is difficult. Measurement error may well compromise its resolution
and result in misinterpretation. For example, our principal measure of
PMD was based on behavioural observation, while residual endogeneity
symptoms were assessed by self-report. Next, fair assessment requires
features being rated at or near the nadir of the episode, for features
may change during an episode of depression. Again, as noted earlier,
medication effects may influence a range of symptoms and behaviours.
Additionally, having argued that symptoms may be invalidly reported
as a consequence of a severe mood state, we risk having it both ways -
examining clinical features of endogeneity as symptoms and, then for the
analyses, regarding the self-report data set as appropriate for compar-
ative analyses against PMD data assessed by behavioural observation.
Finally, and as noted earlier, none of the diagnostic criterion measures
used in the study is, of necessity, valid. As each provided quite differ-
ing prevalence estimates of melancholia within the sample, some or all
have to have a significant error component, so that interpretation of the
hypothesis is clearly influenced by the particular definitional system.
After respecting such caveats, our results offered partial support. Us-
ing the CORE cut-off score of 8 or more to define melancholia, the neces-
128 G. Parker and D. Hadzi-Pavlovic
sary component of our hypothesis was supported reasonably well against
Newcastle and CLINICAL diagnoses, but only modestly for DSM-III-R.
The sufficient component required demonstrating that everyone with
categorised PMD was assigned a diagnosis of melancholia - and was
met reasonably well against DSM allocations but less well by Newcastle
and CLINICAL diagnoses. Failure to confirm the hypothesis categorically
suggested either that no measure of melancholia was valid (thus weak-
ening any predictive potential of CORE scores) or that "melancholia"
requires additional clinical information - above and beyond information
on PMD. So we might view melancholia as having a "core" of psychomo-
tor disturbance and a "mantle" of independent endogeneity symptoms.
We sought then to define both the extent of the additional mantle and
those endogeneity features contributing to it. To the extent that our
three systems accurately define melancholia, the mantle appeared thin.
Thus, after entering the total CORE score as a predictor of melancholia,
the addition of the total SYMPTOM score increased the kappa by 0.01-
0.06 across the three systems. In addition to that dimensional approach
to endogeneity symptoms, we entered each as an individual item (i.e.,
presence vs. absence). In addition to CORE-defmed PMD, only two
symptoms (anhedonia; mood and/or energy being worse in the morn-
ing) made an additional significant contribution to predicting melancho-
lia across all three diagnostic systems, while an additional two (appetite
and/or weight loss; terminal insomnia) made a significant contribution
to DSM-III-R assignment. Thus, the mantle may well comprise only a
few endogeneity symptom items that are independent of psychomotor
disturbance. CORE score data then would appear to be an efficient pre-
dictor of melancholic classification, and the CORE measure appears to
capture the essence of melancholia.
It may well be, of course, that there are heterogeneous expressions
of melancholia (some weighted more to endogeneity symptoms - as in
DSM-III-R - and others weighted more to PMD), with varying aetiolo-
gies, outcomes and clinical features. On the other hand, melancholia may
be a reasonably homogeneous entity with neurobiological underpinnings
variably influencing mood, cognition, movement and other constructs,
and in such a way accounting for some heterogeneity in presentation
across symptom and behavioural clinical domains. If not assisted by the
findings of some "gold standard" such as the Lewy Body in Parkinson's
disease was once considered (but now viewed as less accurate than good
clinical judgment in making a positive diagnosis), pursuit will be less
straightforward. Presumably progress will involve an iterative process
Development and Structure of the CORE System 129
- of "alternately regarding the clinical diagnosis as the independent,
then the dependent variable in relation to other data" (Carroll 1989).
Such other data will presumably involve neurobiological, including neu-
rochemical, neuroendocrinological, neuroimmunological and brain imag-
ing (both functional and structural) studies as detailed in Chapter 3.
The potential importance of the CORE-based system is then in suggest-
ing both a means for estimating the probability of melancholia as well as
giving some insights into those neuropathological processes most likely
implicated in the pathogenesis of the condition. In Part III of this book
we consider a number of candidate neurobiological processes involved in
melancholia, and then return to the potential utility of the CORE system
as a useful strategy for the necessary iterative process of refining the
clinical definition of melancholia.
7
Reliability of the CORE Measure
DUSAN HADZI-PAVLOVIC
GORDON PARKER

Introduction
We provide an overview of two studies examining the reliability of the
CORE II measure in the hands of clinicians (with a comparison of
ratings made by relatives and by psychiatrists having been reported in
Chapter 6). The first (MDU) study was undertaken by the consultant
psychiatrists involved in the development of the CORE measure. There
is a risk to any such endeavour - in that the consultants, as experienced
clinicians, may rate on the basis of clinical intuition (i.e., they might ob-
serve a patient, "smell" melancholia, and "rate up" on the CORE system -
or conversely "rate down" when assessing non-melancholic depression),
bringing about invalid agreement between CORE ratings. The second
(NIMH) study, involving non-MDU staff, provided an opportunity to
overcome any "intellectual incest" or related bias, in addition to allow-
ing the reliability of the CORE system to be tested for non-psychiatrists
and from video (rather than live) interviews.

Methods
The MDU Study. Shortly after designing the CORE II schedule, we
commenced an inter-rater reliability study. The procedure involved one
of five consultant psychiatrists (HB, PB, IH, PM and KW) interviewing
a depressed patient for at least 20 minutes, with two or more of the
other consultants observing the interview. At completion, each consul-
tant made an independent rating of all CORE signs, ratings were reviewed
and disagreements were discussed in detail. If a substantive issue of con-
cern emerged, a descriptor might be modified in light of the discussion,
thus reducing potential or actual ambiguities in the schedule over time.

130
Reliability of the CORE Measure 131
Subjects (all inpatients) were not randomly selected, with the consultant
responsible for the session often choosing a patient posing a diagnostic
dilemma to obtain the benefit of peer review. The final sample involved
35 patients, though not all patients were rated by all five consultant
psychiatrists. Specifically, raters A-E returned 20, 13, 30, 32 and 31
ratings, with two raters present on 5 occasions, three on 12, four on 10,
and five on 8 occasions.

The NIMH Study. As part of the NIMH Collaborative Study of the


Psychobiology of Depression (Katz et al. 1979; Maas et al. 1980), a
number of depressed patients attending six U.S. hospitals were studied
prospectively. As part of the assessment and review, a standardised
videotaped interview was administered (see Katz et al. 1989), with a
split-screen presentation for much of the time capturing both the pa-
tient's face and whole body; thus inter-rater reliability could be assessed.
As the U.S. reliability study has been reported elsewhere (Goldsamt
et al. 1995), we summarise details. Six raters from two New York City
hospitals (Montefiore and Cornell Medical Centers) were recruited, all
having received clinical experience in the U.S. Three were psychology
graduate interns undertaking a clinical research programme, while the
remainder comprised a Ph.D. psychologist, an M.D. research fellow and
a research assistant undertaking a psychology graduate program. All
raters initially viewed the CORE measure training videotape with GP
leading a discussion on rating nuances. The six raters, together with
GP and two senior U.S. raters (Martin Katz, Ph.D., and Peter Stokes,
M.D.), then viewed eight representative videotapes (a "training" sam-
ple) from the NIMH study, with discussion of individual ratings occur-
ring after each. As noted in Chapter 6, several issues raised in such
discussions (especially some nuances in relation to motor agitation and
motor stereotypies) led to immediate modification of CORE instructions
(now included in the instructions given in the Appendix). Following
training, the definitive inter-rater reliability study commenced. The de-
sign involved GP and the six U.S. raters, who evaluated independently
55 randomly selected NIMH videotaped patients, without discussion of
ratings at any study stage.

Reliability Estimates. As customary, each sign was scored on a four-


point scale; a score of 0 indicated that the feature was absent, and scores
of 1-3 represented the sign being present and of increasing severity or
duration. In our analyses, we examined the raters' agreement on:
132 D. Hadzi-Pavlovic and G. Parker
(i) a sign being absent or present (subsequently labelled 0-1)
(ii) severity, by assessing agreement between raters across assignments
(subsequently labelled as 0-3).
In addition, we examined for agreement on total CORE scores. Finally,
using the earlier determined CORE score cut-off of 8 or more for assign-
ment to a melancholic sub-class, we examined agreement between raters
in assigning subjects above and below that cut-off. Inter-rater reliability
was estimated using two principal statistics: the kappa (K) coefficient
for dichotomous data, and the intra-class correlation (ICC) for continu-
ous data such as total scores and ordinal item ratings. Technical details
of the analytic techniques used for the MDU study are provided in an
earlier publication (Hadzi-Pavlovic et al. 1993), while the NIMH study
used similar analytic procedures.

Results
Table 7.1 considers each of the 18 signs (consolidated under the three
separate sub-scales), with mean kappa coefficients of agreement between
raters rating the presence versus absence of each item. The kappas
suggest slight to moderate agreement between raters in evaluating the
presence versus absence of individual items, with the range being 0.26-
0.65 across the MDU sample and 0.19-0.61 across the NIMH sample.
In Table 7.1 we also report intra-class correlations for the individ-
ual items scored 0-3, examining agreement in assessing item severity or
duration. The range of ICCs (0.40-0.79 in the MDU sample) suggests
modest to good agreement there, with a similar range in the NIMH
sample (0.53-0.87) (apart from two agitation sub-scale items: facial ap-
prehension and facial agitation). Importantly, when we propose that the
total CORE score provides an index of melancholia, there was impressive
agreement in total CORE scores, with ICCs ranging from 0.79 to 0.90
across raters in the MDU sample and with a mean ICC of 0.87 in the
MDU study and a mean of 0.82 in the NIMH sample. Intra-class corre-
lations for sub-scale scores were: retardation (0.75 and 0.73, MDU and
NIMH respectively), non-interactiveness (0.84 and 0.76) and agitation
(0.46 and 0.48).
As noted in Chapter 6, analyses identified a cut-off score of 8 or more
in assigning patients to a melancholic sub-class. When we examined
MDU raters' assignment of subjects above or below that cut-off score,
the agreement levels between raters were generally modest (range in
Reliability of the CORE Measure 133

Table 7.1. Inter-rater agreement for signs

Present vs. absentf Severity^


i (ICC)§
MDU NIMH MDU NIMH
Sign sample sample sample sample
Non-interactiveness items
Non-interactiveness 0.41 0.54 0.68 0.75
Postural slumping 0.36 0.35 0.39 0.61
Non-reactivity 0.46 0.58 0.71 0.78
Delay in responding verbally 0.53 0.57 0.71 0.87
Shortened verbal responses 0.62 0.57 0.67 0.80
Poverty of associations 0.57 0.52 0.71 0.82
Delayed motor activity 0.48 0.42 0.63 0.72
Impaired spontaneity of talk 0.61 0.42 0.72 0.80
Slowing of speech rate 0.26 0.42 0.40 0.69
Agitation items
Facial apprehension 0.32 0.24 0.40 0.19
Inattentiveness 0.41 0.42 0.49 0.54
Facial agitation 0.39 0.19 0.47 0.15
Motor agitation 0.65 0.41 0.79 0.64
Verbal stereotypy 0.53 0.61 0.78 0.53
Stereotyped movements 0.60 0.48 0.64 0.74
Retardation items
Facial immobility 0.39 0.50 0.47 0.70
Body immobility 0.53 0.51 0.66 0.78
Slowed movements 0.49 0.37 0.67 0.73
Total CORE score 0.87 0.82
f Ratings coded 0-1. i Ratings coded C
% Coefficient kappa for agreement.
§ Intra-class correlation for agreement.

kappas of 0.27 to 0.77). In the NIMH study, when the average total
CORE score returned by the U.S. raters was compared with the score of
the Australian rater (GP) for the 55 subjects, the two scores were in
agreement in assigning 45 (82%) of the cases as above (n = 21) or below
(n — 24) the cut-off. For patients at either side of the boundary of the
cut-off (e.g., those receiving 7 from one rater and 8 from the other), such
a variation in the total CORE score is, in reality, intrinsically trivial if a
cut-off is not being imposed. When, however, a cut-off score is assigned
we are using a tougher assessment of agreement, and even such trivial
134 D. Hadzi-Pavlovic and G. Parker
differences can result in a kappa lower than might be expected from the
ICC. We therefore examined the effect of creating a "band" rather than
a simple cut-off score. In the MDU sample, we used our latent class-
derived allocation to a melancholic class (see Chapter 6), to select lower
and upper cut-off scores where the probability of being a melancholic was
set at 5% or less and 95% or more respectively - in practice, patients
who scored 6-10. Regarding that as a band of unclear diagnostic status,
we then examined the extent to which one assessor returned a score in
the residual non-melancholic range (< 6), while the other returned a
score in the melancholic range (> 10), or the converse, and established a
disagreement rate of 4.7% (4/85). Narrowing the band, by using cut-off
scores where the probabilities of allocation to the melancholic latent class
were set at 20% or less and 80% or more (i.e., creating a band of scores
7-9), resulted in a disagreement rate of 9.5% (10/105) for one assessor
allocating to the melancholic class and the other to the non-melancholic
class. Imposing a similar band of uncertainty (7-9) to the NIMH data
reduced the level of disagreement about melancholia/non-melancholia
assignment from 18% to 10%, almost identical to the MDU estimate.

A Comparison of Clinician-rated and Family


Corroborative-witness Data
Of the 413 depressed patients in our CORE II study, 141 (34%) pro-
vided us with data (including symptoms and signs) from a family cor-
roborative witness. As reported in Parker et al. (1992a), agreement was
minimal or non-existent between those ratings and interview-based rat-
ings by our clinicians of the symptoms and signs. The corroborative
witness raters were much more likely to rate a sign as both present and
more severe. Such minimal agreement meant that we could not claim
validation of either our interview-based ratings of symptoms or of signs.
The possibility that family members could provide important informa-
tion on signs of psychomotor disturbance (a potential advantage when
the psychiatrist might have no awareness of the patient's base level of
functioning) could not be supported. For whatever reason, it appeared
that family members were much more likely to rate non-existent features
(in the view of the clinician) as present, and also to rate existing features
as more severe.
Reliability of the CORE Measure 135

Discussion
We concede several factors that may have had some influence on reliabil-
ity estimates in the principal inter-rater studies. Non-random selection
of diagnostically difficult patients in the MDU study may have compro-
mised estimates. By contrast, the detailed discussion after each rating
(and some modification of descriptors and decision rules, if important)
may have contributed to improving MDU reliability estimates over time.
Both issues were redressed to some degree in the NIMH study, where
random selection of tapes occurred and where there was no discussion
between raters during the study. Reliance on videotaped interviews
(disallowing the rater from interacting with the patient to explicate the
presence and severity of certain signs) might appear, theoretically, to
be a limitation - but the similar reliability estimates (across the MDU
and NIMH studies) suggest that any such limitation was likely to be
slight, or was compensated by the fact that the standardised NIMH in-
terview required the patient to perform cognitive and motor tasks. The
inter-rater reliability calculations reported for the MDU study provide
a reasonable estimate of an important property of the measure within a
group of trained clinicians. Earlier on we conceded the possibility that
reliability estimates might have been spuriously high as a consequence
of studying only raters who were sophisticated (in the sense of being
involved in the development of the CORE system). Key advantages
to the NIMH study, then, were that raters were not involved in the
development of the CORE measure, they were from another national
setting and a number were non-psychiatrists. The general consistency
of inter-rater reliability estimates across the two studies suggests then
that high reliability (at least for the important total CORE score) can
be obtained after training, that the CORE system is not culture specific
(e.g., to Australian nuances) and that rating does not of necessity require
a psychiatrist. As the NIMH estimates were derived from videotaped
interviews, we can also conclude that adequate reliability in CORE rat-
ings can be achieved from videotaped interviews - subject, of course, to
the videotapes being of adequate quality and using a strategy (such as
a split screen approach) which allows both facial and whole body signs
to be readily observable.
The intra-class coefficients assessing severity of CORE scores and sub-
scale scores were impressive across both studies, with the only caveat
being for agitation sub-scale scores. As noted (in Chapter 6), the two
reliability studies led to some changes to the instructions for rating ag-
136 D. Hadzi-Pavlovic and G. Parker

it at ion items, and it is likely that any subsequent inter-rater reliability


study undertaken now would return higher levels of agreement in rating
agitation items, subject to the new instructions being respected.
The reliability of total CORE scores is, of course, of central importance
in considering its capacity to allow judgements about the probability of
melancholia. The overall (i.e., across all raters) intra-class correlation
coefficients of 0.87 and 0.82 across the two studies are worthy of note,
in offering support for the inter-rater reliability of the CORE system as
an index measure. By contrast, agreement in allocating patients above
or below a total CORE score of 8 (to imputed classes of melancholia and
non-melancholic depression) was less satisfactory. As noted, when sub-
jects receive from different raters total CORE scores which cluster around
the cut-off score (e.g., 8 vs. 7), the estimate of agreement using a cut-off
(and a categorical analysis) can be much lower than the estimate us-
ing the CORE scores without any cut-off (and a dimensional analysis).
Such findings are not surprising. In the absence of perfect agreement,
there will always be some rater error from whatever source and, unless
there is a wide point of rarity, there will be patients who are genuinely
just either side of the cut-off. Given that limitation, it is inevitable
that some real 7s will receive a rating of 7 from one rater and 8 from
another. In response to this well-known problem the question is, both
here in relation to CORE scores and as a general proposition, how seri-
ously one should treat a cut-off (or, indeed, any yes/no assignment). For
some purposes (e.g., comparing groups on mean scores), it is probable
that it does not matter, in that researchers assume that false positive
allocations cancel out false negative allocations. For making treatment
decisions, distinctions between diagnostic black and white groups may
benefit from allowing a grey category, such as one included in the Re-
search Diagnostic Criteria system (Spitzer, Endicott and Robins 1980)
for assigning endogenous depressive patients to probable or definite sub-
groups. Supplementary analyses indicated that by imposing such a band
(i.e., from 7-9), inter-rater disagreement on melancholia assignment was
reduced to 10%, a satisfactory result.
We suggest that findings from the two principal reliability studies
support researchers using the CORE system in the following ways:

(i) dimensionally, akin to the Newcastle endogeneity index system (Car-


ney, Roth and Garside 1965);
(ii) with a single cut-off score imposed; or,
(iii) if there is a need to reduce error emerging from imposing a cut-
Reliability of the CORE Measure 137
off, with a range of scores representing a band of diagnostic uncer-
tainty.
If the CORE system is used as a dimensional measure, any rater's ten-
dency to score high or low should be considered in the study design,
although our ICC data for total CORE scores in the MDU study which
contrasted each rater with each other (see Hadzi-Pavlovic et al. 1993)
suggest that any such response set is slight in trained raters. In some
studies, in addition to training, it may be worth using multiple raters
and relying on the general finding that a composite score (i.e., a mean
across raters) is usually more reliable than a score generated by any one
rater.
Finally, CORE scores returned by family members (when contrasted
with ratings by consultant psychiatrists) indicated that the CORE mea-
sure should not be completed by such corroborative witnesses, and that
it should be completed only by trained clinical staff.

Conclusion
We suggest that, in the hands of a trained clinician (psychiatrist or
other), the CORE system has been established as providing a reliable
estimate of psychomotor disturbance. As for other measures, such as the
Hamilton (1960) scale, training against benchmark standards or guide-
lines is necessary to appreciate the nuances of the constructs as well as
to help standardise ratings and reduce inter-rater variance.
8
Validity of the CORE:
I. A Neuroendocrinological Strategy
PHILIP MITCHELL

Introduction
In a recent review, Rush and Weissenburger (1994) concluded: "Further
research is needed to empirically test the biological and psychological
features associated with melancholic depression." In this chapter, we
examine the dexamethasone suppression test (DST) as a biological val-
idator of melancholia as defined by the CORE system. In Chapter 6 we
demonstrated quite strong differentiation by DST non-suppression be-
tween those assigned by the CORE I system as melancholic and those
assigned as non-melancholic, and briefly reported similar differentiation
in a sample of the CORE II subjects. Here we also examine DST dif-
ferentiation in a sample of CORE II subjects, complemented with ad-
ditional recruitment (beyond closure of that previous study) to ensure a
large sample size.
Previous attempts to validate definitions of melancholia (or endogen-
ous depression) against DST results have had mixed outcomes. Whereas
the DST appears to discriminate reasonably well between the depressive
sub-types defined by the Newcastle Diagnostic Index (Carney, Roth and
Garside 1965; Coppen et al. 1983; Holden 1983; Georgotas et al. 1987;
Zimmerman et al. 1986c; Staner et al. 1992), there has been much less
consistency in its capacity to differentiate RDC and DSM-m (Davidson
et al. 1984; Philipp, Maier and Holsboer 1986; Rush and Weissenburger
1994) sub-groups.
Reflecting on the strategy of validating the DST as a marker for melan-
cholia, Zimmerman et al. (1986a) commented: "How does one validate a
biological marker of endogenous depression when a valid clinical defini-
tion does not exist?" They proceeded by testing the validity of the DST
against the seven symptoms identified by Nelson and Charney (1981)

138
A Neuroendocrinological Strategy 139
as characteristics of endogenous depression. Of these, only scores for
psychomotor retardation and depressed mood were greater in the non-
suppressors, a finding consistent with results of most studies which have
examined the relationship between the DST and psychomotor distur-
bance (Brown et al. 1988; Calloway et al. 1984; Klein et al. 1984).
Zimmerman et al. (1986a) also observed that non-suppressors more
frequently appeared depressed (as opposed to merely reporting feeling de-
pressed), and suggested, "Observational symptom ratings may be more
valid than interview based ratings, therefore possibly accounting for our
failure to find a significant association with anhedonia and reactivity."
As an observational instrument, the CORE system therefore provides
such a methodology.
In addition to so attempting to validate the CORE system, we ex-
amine its comparative DST discrimination against two other measures
of melancholia (DSM-III-R and Newcastle). We both examine formally
assigned classes, as well as examine for any dimensional relationship
between DST findings and Newcastle and CORE scores, with the latter
approach rarely used in previous DST studies. Also, as age, dexam-
ethasone levels and weight loss may affect both cortisol levels and DST
findings, we examine their influences as potential confounding variables,
a refinement used in few studies examining the DST as a biological val-
idator of melancholic depression.

Methods
Consecutive inpatients with a primary major depressive episode were
included in the study, with (as noted) recruitment continuing after the
CORE II study had ended to ensure a large sample. After applying
standard exclusion criteria (Carroll et al. 1981a), 114 patients were re-
cruited, excluding those on anticonvulsants, but including those taking
antidepressant and antipsychotic medications as these do not appear to
affect DST findings (e.g., Klein et al. 1984; Hunt, Johnson and Caterson
1989). Fourteen subjects were subsequently excluded, as their cortisol
and dexamethasone results for each post-dexamethasone time point were
not available, leaving a final sample of 100 subjects.
In addition to DSM-m-R assignment, patients were assigned to the
Newcastle endogenous category if they scored 6 or more on that scale,
and as melancholic by the CORE system if they had a CORE score of 8
or more. Weight loss for the current episode was rated as nil, slight (<
3 kg), moderate (3-5 kg) or marked (> 5 kg).
140 P. Mitchell

Blood samples for pre-dexamethasone cortisol concentrations were


taken at 1600 h on day one, with 1 mg dexamethasone being admin-
istered orally at 2300 h that night. The following day, blood was col-
lected at 0800 h and 1600 h for cortisol and dexamethasone determina-
tions, and at 2300 h for cortisol levels only. Cortisol non-suppression
was defined by cortisol concentrations > 138 nmol/litre (5 mg/dl) at
0800 h, 1600 h or 2300 h following dexamethasone administration (Car-
roll et al. 1981a), and results are expressed as means (db standard devia-
tion). Post-dexamethasone cortisol concentrations were examined both
dimensionally and categorically, with the latter being suppression versus
non-suppression.

Results
The final sample comprised 69 females and 31 males, with a mean (±SD)
age of 54.4 (±17.5) years. The mean Hamilton score was 23.2 (±6.5) for
the 17-item measure, and 25.1 (±7.1) for the 21-item measure. Quite
varying rates of patients were diagnosed as melancholic or endogenous
according to the various systems (51% by Newcastle; 59% by CORE;
and 77% by DSM-III-R), and we now use the term "melancholia" to
subsume both melancholic and endogenous depression assignment for
relevant subjects assigned by each of the three systems. Both CORE and
Newcastle-assigned melancholic patients were significantly older than
non-melancholic patients (60.5 vs. 45.2; 62.4 vs. 46.1 years respectively),
while the same trend was not significant for DSM-III-R assignment (56.1
vs. 48.6).

Cortisol Concentrations. There were no significant differences in cor-


tisol concentrations between melancholic and non-melancholic patients
as allocated by the three diagnostic systems for the day one 1600 h cor-
tisol concentrations and, consequently, these levels will not be consid-
ered further. The 0800 h day two cortisol concentrations were signifi-
cantly higher in both Newcastle- and CORE-, but not DSM-iil-R-defined
melancholies. The 1600 h cortisol levels were higher in melancholies de-
fined by all systems. The 2300 h concentrations were higher only in the
Newcastle-defined melancholies.

Non-suppression Rates in Melancholic and Non-melancholic


Patients. Table 8.1 shows that the non-suppression rates were dis-
tinctly higher for Newcastle- and CORE-assigned melancholic patients
A Neuroendocrinological Strategy 141

Table 8.1. Non-suppression rates (%) in variably assigned melancholic


(MEL) and non-melancholic (nMEL) depression

DSM-III-R Newcastle CORE


Time (h) MEL nMEL MEL nMEL MEL nMEL
0800 26.0 13.0 33.3 12.2* 33.9 7.5**
1600 46.8 26.1 58.8 24.5** 52.5 27.5*
2300 28.6 17.4 35.3 16.3* 33.9 15.0*
0800/1600 51.9 26.1* 62.7 28.6*** 57.6 30.0**
0800/1600/2300 53.2 34.8 62.7 34.7** 57.6 37.5*

2 x 2 chi-squared statistics: *P < 0.05; **P < 0.01; ***P < 0.001.

at each of the three time points individually or when combined (0800 h


and/or 1600 h; 0800 h and/or 1600 h and/or 2300 h). For DSM-III-Re-
assigned melancholic patients, non-suppression rates were higher only in
the combined 0800 h and/or 1600 h analysis.

Dexamethasone Concentrations. There were no significant differ-


ences in dexamethasone concentrations between DSM-in-R-allocated mel-
ancholic and non-melancholic patients. However, against both Newcas-
tle and CORE allocations, dexamethasone concentrations were lower in
melancholic patients, and with differences more distinct in the 1600 h
analyses.
(As there were no substantive differences in either cortisol or dexam-
ethasone concentrations between DSM-m-R-allocated melancholic and
non-melancholic sub-groups, subsequent analyses of the effects of poten-
tial confounding variables were restricted to the CORE and Newcastle
systems.)

Age, Cortisol and Dexamethasone Influences. Age correlated sig-


nificantly with both Newcastle (+0.54) and CORE (4-0.46) scores and
with both 0800 h (+0.34) and 1600 h (+0.35) cortisol concentrations, but
not with either 0800 h or 1600 h dexamethasone concentrations. Corti-
sol concentrations correlated significantly with Newcastle scores at both
0800 h (+0.28) and 1600 h (+0.35), and also with CORE scores at 0800 h
(+0.37) and 1600 h (+0.39). In view of this apparent association be-
tween cortisol concentrations and the CORE score, we then examined
for any specific relationship between cortisol levels and CORE sub-scales.
Both 0800 h and 1600 h cortisol concentrations correlated significantly
142 P. Mitchell

and similarly with each of these CORE sub-scales (i.e., +0.30 and +0.39
for retardation; +0.37 and +0.34 for agitation; +0.34 and +0.33 for
non-interactiveness).
As with the cortisol levels, dexamethasone concentrations correlated
significantly (and negatively) with both Newcastle and CORE scores at
0800 h and 1600 h. Examined against CORE sub-scales, 0800 h dexa-
methasone concentrations correlated significantly only with non-inter-
activeness (—0.25), while 1600h levels correlated significantly with re-
tardation (^0.33), agitation (—0.31) and non-interactiveness (—0.39).
As demonstrated in previous studies, there were significant negative
correlations between cortisol and dexamethasone concentrations at both
0800 h and 1600 h, with correlation coefficients ranging from —0.33 to
—0.43. There were, however, no significant correlations between cortisol
or dexamethasone concentrations and Hamilton scores.

The Effect of Weight Loss. As previous studies have suggested a


relationship between rates of non-suppression and degree of weight loss,
rates of weight loss were compared in suppressors and non-suppressors.
No significant differences were demonstrated at either the individual or
combined time points, indicating that weight loss was not a significant
factor in determining non-supression rates in this sample.
When cortisol concentrations in depressed patients with moderate/
marked weight loss were compared to those in depressed patients with
no/slight weight loss, levels were significantly higher in the former at
0800h (156.2 vs. 91.5), but there was no significant difference at 1600h
(148.6 vs. 163.1). Dexamethasone concentrations, however, were signifi-
cantly lower at both 0800 h and 1600 h in those with moderate/marked
weight loss (i.e., 7.4 vs. 9.6; and 2.3 vs. 3.2) for respective time points.
As we had previously demonstrated relationships between dexametha-
sone concentrations and CORE and Newcastle scores, we examined whe-
ther weight loss was associated with CORE and Newcastle scores. Both
CORE and Newcastle scores were higher in those with moderate/marked
weight loss (CORE: 16.5 vs. 10.6; Newcastle: 6.1 vs. 3.5). These findings
indicate that weight loss was significantly related to CORE and Newcastle
scores.

Cortisol Concentrations: The Effect of Age and Dexametha-


sone Concentrations. Having demonstrated that cortisol concentra-
tions were significantly linked with age and dexamethasone levels, the ef-
fects of the latter on cortisol levels between Newcastle and coRE-defined
A Neuroendocrinological Strategy 143
melancholic and non-melancholic patients were examined using analyses
of variance. For Newcastle-defined groups, differences in cortisol con-
centrations were no longer significant after covarying for age at both
0800 h and 1600 h. After covarying for dexamethasone concentration,
the differences in cortisol levels remained significant only at 1600 h. For
CORE-defined depressive sub-groups, differences in cortisol levels at both
0800 h and 1600 h also lost significance after covarying for age. However,
after covarying for dexamethasone concentrations, the difference in cor-
tisol levels remained significant at 0800 h, and there was still a trend
towards significance at 1600 h.

Non-suppression Rates: The Effect of Age and Dexamethasone


Concentrations. Effects of these covariates on non-suppression rates
were examined using logistic regression analyses. For Newcastle-defined
sub-groups, differences in non-suppression rates at the individual 0800 h
and 1600 h time points were no longer significant after partialling out
age. However, when non-suppression at the combined 0800 h and/or
1600 h time points was considered, there was still a significant difference
between melancholic and non-melancholic groups. When relevant dex-
amethasone concentrations were partialled out, the differences between
groups were no longer significant at either 0800 h or 1600 h.
For CORE-defined groups, differences were no longer significant at
0800 h or 1600 h after partialling out the effect of age. When non-
suppression at the combined 0800 h or 1600 h time point was consid-
ered, there was similarly a loss of significance after partialling out age.
When the effect of dexamethasone concentrations was partialled out,
significant differences persisted at 0800 h but not at 1600 h.

Relationships Between Non-suppression Rates and Newcastle


or CORE. Associations between scores on the CORE or Newcastle mea-
sures and non-suppression rates can be regarded as having dose-response
relationships. Scores on each measure were therefore divided into a
number of ranges, and scores falling into each range were grouped and
assigned the mid-point of the range as their value of the dose.
Figure 8.1 shows a virtually linear relationship between increasing
CORE scores and DST non-suppression rates; while a linear relationship
between Newcastle and non-suppression rates is less clearly evident. The
non-suppression rates in those with severe CORE-defined psychomotor
disturbance are worthy of emphasis. Thus, for those with minimal CORE
scores, the DST non-suppression rate was approximately 30%, for those
144 P. Mitchell

with moderate CORE scores it was more than 50%, while for those with
the highest CORE scores, DST non-suppression was approximately 90%.
Relationships between dose and the proportion of patients non-sup-
pressing in each group were then examined using logistic regression, with
the principal predictor variable being (in separate analyses) CORE and
Newcastle scores, and with the outcome variable being non-suppression
at the combined 0800 h or 1600 h time points, with differing models
also testing the added influences of age and weight loss as predictors.
The null model assumed a constant rate of non-suppression. Models
examining the fit of the measures alone were significant (CORE: P <
0.001, Newcastle: P — 0.01), indicating that as scores increased on the
measures, non-suppression rates increased. Age alone was a significant
predictor, while the addition of weight loss to age had no effect. The
addition of both CORE and Newcastle scores (in separate models) to
the age and weight predictors was significant (P — 0.003 and P = 0.01
respectively), indicating that as scores on the CORE and Newcastle scales
increased, non-suppression rates increased irrespective of age and weight
loss.

Discussion
Attempts to demonstrate associations between DST findings and various
definitions of melancholia have produced variable findings in the past.
Our finding of a lack of any association between the DST and the DSM-
III-R definition of melancholia is consistent with the negative findings
reported against DSM-m definition (Davidson et al. 1984; Philipp et al.
1986; Rush and Weissenburger 1994). Similarly, our finding of an asso-
ciation between the DST and the Newcastle definition of melancholia is
consistent with a number of previous reports (Coppen et al. 1983; Zim-
merman et al. 1986c; Zimmerman et al. 1986a; Rush and Weissenburger
1994). Our demonstration of significant associations between DST non-
suppression and CORE scores is consistent with an extensive literature
indicating that psychomotor disturbance - usually retardation but also
agitation (Calloway et al. 1984; Zimmerman et al. 1986a) - is the most
replicable clinical correlate of DST non-suppression.
We also found significant dose-response relationships for both New-
castle and CORE scores against non-suppression rates. Coppen et al.
(1983) have previously described a linear relationship with the Newcas-
tle scale, while Zimmerman et al. (1986c) found a non-linear relation-
ship. Our finding of links between post-dexamethasone cortisol levels
A Neuroendocrinological Strategy 145

Observed N= 8
vsupprt

90- Expected under logistic model


80- N= 10

1 70-

i 60- N= 13
CD
50-

s>
s
40-
N= 30
30-
20-
CL
10-
0-

2 7 12 17 22 27 32

Mid-points of Grouped CORE Scores

N= 5
100-
Observed
90- Expected under logistic model

80

60-

50-

40-

30-

20-

10-

0-

-1.5 0.5 2.5 4.5 6.5 8.5 10.5

Mid-points of Grouped Newcastle Scores

Fig. 8.1. Logistic Model for Probability of Non-suppression as a Function


of CORE Score. Figure Shows Proportion of Patients Expected to be Non-
suppressors under the Model and the Proportion of Patients Observed to
Non-suppress.
146 p. Mitchell

and both Newcastle and CORE scores is also compatible with the lit-
erature. A significant positive correlation between the Newcastle scale
and cortisol levels has been previously reported by Coppen et al. (1983)
and Christensen et al. (1986). The significant correlation between CORE
and post-dexamethasone cortisol levels is consistent with the reports of
Klein et al. (1984) and Smith et al. (1988), who both found a significant
correlation between post-dexamethasone cortisol levels and other retar-
dation scales. Staner et al. (1992), using stepwise multiple regression,
also reported that the Newcastle Scale items for psychomotor retarda-
tion and weight loss contributed most to the variance of the 2300 h cor-
tisol levels. Although other studies have suggested that agitation may
be more related to hypercortisolaemia than to retardation (e.g., Brown
et al. 1988; Meador-Woodruff et al. 1990; Miller and Nelson 1987), we
established similar correlations with each of the agitation, retardation
and non-interactiveness CORE scales.
Unlike most previous investigations examining for associations be-
tween post-dexamethasone cortisol concentrations and depressive sub-
groups, we examined for the effects of age, dexamethasone concentra-
tions and weight loss. Weight loss was not found to be strongly asso-
ciated with either suppressor status or cortisol levels, and will not be
considered further. When we considered melancholia as defined cate-
gorically by either the CORE or Newcastle system, the associations with
either cortisol levels or DST non-suppression were no longer significant
after partialling out the effect of age, indicating the age dependence
of this relationship. However, when the CORE system was considered
dimensionally, the correlation between CORE scores and both cortisol
concentrations and DST non-suppression rates remained highly signif-
icant after partialling out the effect of age, indicating a residual age-
independent relationship. By contrast, correlations between the Newcas-
tle Scale and cortisol concentrations were either insignificant (at 0800 h)
or barely significant (at 1600 h) after accounting for the effect of age.
These findings indicate that there is an important age-independent pos-
itive relationship between the CORE rating for psychomotor disturbance
and hypothalamic-pituitary-adrenal axis (HPA) over-activity, which is
demonstrated more clearly when CORE scores are considered dimension-
ally rather than categorically.
Few previous DST studies have examined for the effect of dexa-
methasone levels (Maguire et al. 1987), despite clear associations being
reported between dexamethasone concentrations and the prevalence of
non-suppression (e.g., Arana, Workman and Baldessarini 1984; Berger
A Neuroendocrinological Strategy 147
et al. 1984; Johnson et al. 1984; Holsboer et al. 1984). When we as-
signed subjects categorically by the CORE or Newcastle system, the ef-
fect of partialling out dexamethasone concentrations led either to a loss
of significance or to a reduction in the level of significance dramatically.
However, when the associations between cortisol and CORE or Newcastle
scores were considered dimensionally, correlations remained significant,
and distinctly so for the CORE system. The latter finding is consistent
with a previous study, where Smith et al. (1988) reported that the cor-
relation between cortisol levels and psychomotor retardation persisted
after controlling for dexamethasone levels.
Thus, in general terms, the associations between both CORE and New-
castle scores and DST findings persisted after partialling out either age
or dexamethasone concentrations when these systems were considered
dimensionally, but not when they were examined categorically. This
indicates the likelihood of greater utility of dimensional over categori-
cal diagnostic scores in examining HPA axis activity in depression - a
conclusion also posited by Maes et al. (1990).
In relation to DST differentiation, the CORE system appeared far more
differentiating than DSM-III-R in particular, but also more differentiat-
ing than the Newcastle index, suggesting the centrality of psychomotor
disturbance.
The other major findings of this study were:
(i) the lower dexamethasone concentrations in Newcastle- and CORE-
defined melancholic patients (though not in DSM-iii-R-defined melan-
cholia, akin to the negative findings of Maes et al. (1990) with DSM-
III; and
(ii) the negative correlation between dexamethasone levels and dimen-
sional CORE and Newcastle scale scores.
The inverse relationship between psychomotor disturbance (assessed by
CORE scores) and dexamethasone concentration is intriguing. Possible
explanations include depressed patients with greater psychomotor dis-
turbance either metabolizing dexamethasone more rapidly or absorbing
it more slowly. Delayed absorption is an attractive proposition, as con-
stipation (indicative of reduced gastrointestinal motility) has been clas-
sically associated with overt psychomotor retardation (see Chapter 12).
Reduced gastrointestinal motility could lead to impaired absorption of
dexamethasone. A few quality research studies of dexamethasone phar-
macokinetics in depression (e.g., Maguire et al. 1990; Holsboer et al.
1986b; Holsboer, Wiedemann and Boll 1986a; Guthrie 1991) suggest,
148 P. Mitchell

however, that impaired absorption is unlikely, and that an abnormality


of dexamethasone metabolism is more likely.
Such findings suggest that high CORE (and Newcastle} scores are likely
to be associated with an increased metabolism of dexamethasone. Possi-
ble mechanisms are currently unclear, and the links warrant further ex-
amination. The importance of these links is also suggested from a recent
report of Devanand et al. (1991), who found that clinical recovery during
a course of ECT correlated more closely with normalization of dexam-
ethasone (from low levels) than of cortisol concentrations. Such a finding
indicates that the central disturbance in melancholia may link with dex-
amethasone metabolism, as well as with cortisol over-production.
In addition to examining the validity of the CORE system, the cur-
rent study has demonstrated two potential novel determinants of HPA
activity in depression. First, the dimensional CORE (and to a lesser ex-
tent Newcastle) ratings of psychomotor disturbance correlated positively
with post-dexamethasone cortisol concentrations independently of age,
weight loss and dexamethasone levels. Secondly, both CORE and New-
castle scores correlated negatively, and again independent of age, with
dexamethasone concentrations themselves, indicating a central impair-
ment of either the metabolism or absorption of dexamethasone in those
with melancholia identified by those two systems.
Finally, to the extent that melancholia is a disorder associated with
DST non-suppression, analyses offer strong validation of the CORE sys-
tem as a measure of melancholia by such a biological marker. How-
ever, the strong linear relationship between CORE scores and DST non-
suppression (with DST non-suppression rates rising from a base of 30%-
90% as CORE scores increased) indicates that DST non-suppression may
relate more to the level of psychomotor disturbance than to melancholia
as an entity.
9
Validity of the CORE:
II. Neuropsychological Tests
IAN HICKIE

Neuropsychological Testing in Depression


Although extensively researched (see Watts 1992; Austin et al. 1992b),
neuropsychological testing in patients with depressive disorders contin-
ues to reveal disparate findings, though clear evidence of impairment
across a range of functions has been documented in those patients with
more severe and/or psychotic disorders, those hospitalized for treatment
and/or those receiving psychotropic medication (Robbins, Joyce and
Sahakian 1992). The deficits described have primarily involved the do-
mains of attention, concentration and motor speed (the so-called "sub-
cortical" tasks). Various aspects of memory and planning have also
been noted to be affected, though the nature of such impairments has
often been described in non-uniform terms, reflecting the specific cog-
nitive paradigm under investigation. In the interpretation of memory
impairment, emphasis continues to be placed on the possible roles of
effort, motivation and attention to the task, all of which are impaired
in depressed patients. However, memory deficits that do not reflect
such factors are probably a feature of most depressive disorders (Austin
et al. 1992b). Aphasias and apraxias (cortical tasks), have not generally
been associated with depressive disorders and, when present, have typ-
ically been considered as positive evidence of an underlying dementing
disorder.
Differential patterns of neurocognitive impairment may be evident in
less severe forms of depression (i.e., greatest impairment in subcortical
rather than cortical tasks). It has been suggested that the most severe
forms of depression are also associated with more significant memory
and frontal deficits (Austin et al. 1992b). It had generally been assumed
that the cognitive deficits associated with depression, even when such

149
150 /. Hickie
cognitive deficits are severe, are reversible with effective treatment of the
depressed state (thereby, giving rise to the notion of pseudo-dementia
secondary to severe depression or other neuropsychiatric disorders [Kiloh
1961]). More recent longitudinal studies have challenged this assumption
(Alexopoulos et al. 1993a), indicating that severe depressions, particu-
larly those occurring later in life and presenting with concurrent cogni-
tive impairment, may be associated with irreversible cognitive deficits.
In these cases, it is likely that such depressions represent the early stages
of irreversible degenerative disorders.
Questions remain as to the specificity of any neuropsychological find-
ings in depression, and whether any of the abnormalities described reflect
anything more than positive correlations between increasing levels of
symptomatology (that is, as the number of symptoms increases, so does
the patient's chance of displaying each of the individual depressive symp-
toms, including cognitive symptoms). Robbins et al. (1992) concluded
that the cognitive disturbance encountered in patients with depression
showed features of both the subcortical and cortical dementia syndromes
but, importantly, also showed marked differences. They proposed that
depression should be considered as having "its own distinctive cognitive
profile." This conclusion is particularly relevant to current neurobiolog-
ical researchers who (i) seek to link findings from neuropsychology with
those from structural and functional neuroimaging studies; and (ii) seek
to compare and contrast findings, within these two modalities, between
depressed patients and relevant neurological control groups, particularly
those disorders affecting the basal ganglia (e.g., Parkinson's disease) or
deep white matter structures (e.g., multiple sclerosis).

Utility of Neuropsychological Testing


Neuropsychological test performance can be conceptualised as a sophis-
ticated and standardised form of clinical assessment, where observa-
tions have been narrowed simply to include only cognitive tasks. It
has distinct advantages empirically when compared with other forms
of behavioural assessment, as it is relatively independent of individual-
operator differences and has acceptable inter-rater and test-retest relia-
bility. Results from such standardised tests can be compared and con-
trasted with other forms of clinical assessment (e.g., symptom reports,
mental status findings, standardised scales), and may provide indepen-
dent evidence for the validity of other key clinical findings. For example,
in melancholia, objective slowing on electronic measures of reaction time
Neuropsychological Tests 151
(or other timed tasks) may validate the clinical notion that patients with
a particular depressive sub-type (melancholia) are characterised by psy-
chomotor slowing.
Neuropsychological testing, especially when performed longitudinally,
may provide insights into possible aetiological processes (e.g., irreversible
degenerative vs. state-dependent pathologies), and the pattern of distur-
bance recorded can also be used to make inferences about the possible
anatomical sites (or systems) involved within the central nervous system
(CNS). Hence the subcortical versus cortical differentiation described
above, as well as the distinctions between frontal, temporal and parietal
cognitive tasks, and between dominant and non-dominant hemisphere
function, may be particularly relevant to studies which seek to corre-
late neuropsychological task performance with functional and structural
neuroimaging.
The bodies of literature describing post-stroke depressive syndromes,
and depressive disorders due to other known disease processes involving
localized areas of the central nervous system, have also been of con-
siderable relevance to the study of relationships between specific cog-
nitive deficits and depressive syndromes. Results from such disorders
are typically considered naturalistic brain lesioning experiments, and al-
low potential correlation of the prevalence and extent of mood disorder,
cognitive impairments and functional disability. Mood disturbance has
generally appeared to be dependent on the site of the lesion (with left
frontal lesions appearing to confer the greatest risk) rather than the
extent of functional disability secondary to the stroke. Comparisons
between the style of mood and cognitive disorder encountered in such
settings with those encountered in primary depressive disorders is now
an attractive area of research. Given the capacity to localize the anatom-
ical site (and, with functional imaging, some of the neurochemical effects
on surrounding regions) of dysfunction in these secondary depressive dis-
orders, important insights into the primary affective disorders may be
gained. For melancholia researchers, the key issues are which types of
secondary depression result in concurrent mood disturbance, movement
disorder and subcortical cognitive impairment. Additionally, whether
any such secondary disorders result in syndromes comparable to psych-
otic depression needs to be closely investigated.
152 /. Hickie

Neuropsychological Studies Performed in Association with


the CORE System
In association with development of the CORE II measure, a limited
battery of neuropsychological tests was performed by trained personnel,
and testing was conducted blind to the clinician's concurrent rating of
CORE scores. The battery focussed on measures of psychomotor speed
(emphasising subcortical tasks) but also included a small range of lan-
guage and memory tasks (emphasising cortical and frontal tasks). Our
goal was to utilise standardised and objective measures of psychomo-
tor function and broader cortical functions so that these data could be
compared with the CORE ratings from the same patients.
Two hundred and thirty-two subjects (comprising both inpatients and
outpatients attending the MDU clinic) were assessed. Subjects covered
the full range of depressive severity and ages encountered in treatment
settings, with case selection extending from patients with first presen-
tations for psychiatric assessment right through to inpatients assessed
just prior to ECT. Results from the initial 78 patients, and the positive
relationships between CORE scores and impaired reaction times, have
been reported previously (Parker et al. 1994). Medication status was
noted and continued throughout the period of testing. The battery of
tests included

(i) a simple reaction time (RT) measure (Huppert 1987) where the
subject is required to press a button after presentation of a single
digit (i.e., 0);
(ii) a more complex decision time (DT) task using the same device
where the subject is required to press a specific button correspond-
ing to the digit presented (i.e., 1-4);
(iii) the trail-making test (TMT), part A (Reitan 1958);
(iv) the symbol-digit modalities test (SDMT; Smith 1968);
(v) the paired associate learning (PALT) subtest from the Weschler
Adult Intelligence Scale (Wechsler 1987); and
(vi) the FAS word fluency task (Spreen and Benton 1969).

Our general experience with this battery is that it can be performed by


the majority of even severely depressed patients, though performance
does decline with age and increasing depression severity.
The relevant clinical data collected included the 21-item Hamilton
Rating Scale for Depression (HRSD); the revised CORE rating system;
and a composite of 18 depressive symptoms (SYMPTOM) which have
Neuropsychological Tests 153
characteristically been associated with the concept of melancholia or en-
dogenous depression over time. Thus, we collected data describing in
standardised form both depression severity (HRSD, SYMPTOM) and psy-
chomotor change (CORE scores), as well as concurrent standardised neu-
ropsychological and electronic measures of psychomotor change. Such
data allowed testing of key hypotheses: first, the validity of the CORE
system; second, the relationship between CORE scores and depression
severity; and third, relationships between age and psychomotor dys-
function, and between age and neurocognitive performance.
The following hypotheses were proposed:
(i) that CORE scores would be strongly correlated with reduced psy-
chomotor speed as assessed by a variety of standardised neuropsy-
chological tests
(ii) that the distribution of neuropsychological test scores would mirror
that of CORE scores (with both indicating bimodality - suggest-
ing the presence of more than one sub-class of patients), and that
such distributions would differ fundamentally from the distribution
of HRSD and SYMPTOM scores (with the latter presenting typical
"unimodal" distributions)
(iii) that neuropsychological tests of "cortical" function would demon-
strate only modest associations with CORE scores, but that such cor-
tical tasks would demonstrate more robust associations with global
depression severity
(iv) that neither depression severity per se nor increasing age would
account completely for associations between CORE scores and di-
minished psychomotor speed.
A sub-set (N = 26) of the patients who were tested neuropsychologi-
cally also underwent magnetic resonance brain imaging (MRI). Thus, we
were able to correlate the degree of impairment on "cortical" and "sub-
cortical" tasks with the extensiveness of changes in subcortical white
matter structures and the basal ganglia (Hickie et al. 1995). Extensive
regions of hyperintensity on MRI have been reported in the deep white
matter and basal ganglia of patients with primary depressive disorders
(Krishnan 1993b; Brown et al. 1992). Such patients, however, typically
have late-onset depressive disorders, severe and often treatment-resistant
disorders, concurrent risk factors to cerebrovascular disease, an absence
of a family history of early-onset affective disorder and, possibly, specific
apolipoprotein E profiles, which have been associated with increased risk
of vascular disease and Alzheimer's disease (Krishnan et al. 1994). These
154 /. Hickie

results suggest possible pathophysiological processes (microvascular dis-


ease) and a resultant neuroanatomical substrate for at least some of the
severe, chronic and disabling depressive disorders of late life (Krishnan
1993b; Hickie et al. 1995).
In our initial analyses of the relationships between our limited neu-
ropsychological battery and CORE scores in 78 patients (Parker et al.
1994), we noted strong associations between total CORE scores and pro-
longed reaction times (for RT: r = 0.74, DT: r = 0.64; TMT: r = 0.57;
all P < 0.001), supporting the validity of the CORE scale as an appropri-
ate measure of psychomotor retardation. In a series of multiple regres-
sion analyses the associations for RT and DT could not be accounted for
simply by increasing age. TMT results appeared to be contributed to by
both age and CORE scores. Patients with melancholia (diagnosed by a
series of available systems, including CORE) had significantly longer RTs,
DTs and TMTs than non-melancholic patients (of the order of x 1.5-2
longer duration), while non-melancholic patients returned reaction times
comparable with those reported for age-matched subjects tested in com-
munity settings.
For the completed series (N = 232), the mean HRSD score was 23.2
(SD 7.9), and the mean CORE score was 11.3 (SD 10.7), indicating the
group was characterised by at least moderately severe depressive dis-
orders (especially given our group's practice of rating Hamilton scores
conservatively). The overall correlation between CORE score and reac-
tion time measures remained strong (RT: r = 0.58, DT: r = 0.57; TMT
r = 0.48, SDMT r = -0.58; all P < 0.001). The correlation for the FAS
task was of the same order (r = —0.52, P < 0.001). The composite mea-
sure of symptoms alone (SYMPTOM) demonstrated correlations with the
measures of reaction time which were in the same direction, but always
of lower magnitudes (for RT: r = 0.37, DT: r = 0.39; TMT r = 0.31;
SDMT: r = -0.41; and for FAS, r = -0.28), suggesting that symptom
reports tap a wider variety of depressive phenomena than psychomotor
change.
Of greater interest than the simple correlations between measures de-
signed to rate psychomotor change were the distributions of reaction
time scores, CORE scores, SYMPTOM scores and HRSD. That is, within
the one patient cohort, we were able to test assumptions concerning
the distribution of the various depressive constructs (see Chapter 3).
Proponents of the unitary model of depressive disorders would predict
that, for global depression severity, and for its component parts (mood,
vegetative symptoms, psychomotor change, cognitive impairment, psy-
Neuropsychological Tests 155
chosis), a unimodal distribution of symptoms and/or constructs should
be apparent. Proponents of the binary model (or any categorical model
hypothesising the existence of distinct depressive sub-types) would argue
that key clinical phenomena (e.g., psychomotor change, psychosis and
various aspects of cognitive performance measuring psychomotor speed)
would demonstrate a bimodal distribution pattern if any such features
or phenomena had specificity to only one depressive type.
The relevant panels in Figures 9.1 and 9.2 demonstrate clearly that
while measures of undifferentiated global severity (HRSD) and symptom
severity (SYMPTOM) present classical unimodal distributions, those rep-
resenting psychomotor speed (CORE scores, RT and DT) do not fit such a
pattern. For each distribution formal mixture analyses were conducted
to test whether a mixture of two populations provided a significantly
better fit to the data that just a single population. The null hypoth-
esis of only one population was rejected for CORE scores (x2 = 87.4,
P < 0.001), for RT (x2 = 52.9, P < 0.001) and for TMT (x2 = 133.3,
P < 0.001), but not for the HRSD scores (x2 = 4.1, P < 0.39), or the
2
SYMPTOM scores (x = 6.9, P < 0.14).
In the MRI phase of the investigation (Hickie et al. 1995), it was
clear that for tasks assessing psychomotor speed, performance declined
as the extensiveness of subcortical white matter lesions, but not grey
nuclei lesions, increased (e.g., for RT: r = 0.54, P < 0.01; and for TMT:
r = 0.43, P < 0.05). No specific associations, however, were noted
between the word-generation (FAS) or memory (PALT) tasks and these
white matter lesions.

Discussion
Given the independence of the neuropsychological data from the clinical
CORE ratings, but the similarity in findings (i.e., strong correlation be-
tween the cognitive and motoric elements, relative independence of such
correlations from severity of depression and age effects) the neuropsycho-
logical data provide the strongest direct validation of the CORE system.
The pattern of distribution of scores and the relatively normal scores of
patients with non-melancholic disorders supports the argument that a
group of patients with psychomotor change and demonstrable neuropsy-
chological impairment (of the slowed reaction time type) constitute a
specific and distinct depressive sub-type. With the demonstration of
strong correlations between reaction-time tests and CORE scores, the
data support the view that a subcortical model of cognitive impair-
156 /. Hickie

S n
M = 39.5 SD = 12.0 Prevalence = 60%
M = 115.2 SD = 72.8 Prevalence = 40%
Estimated total distribution

27.3 68.5 109.8 151 192.3 233.5 274.7 316 357.2 398.5 439.7

Scores on Trails B

8 i
M = 686.7 SD = 126.4 Prevalence = 65%
M = 1181.2 SD = 268.0 Prevalence = 35%
Estimated total distribution

504.8 632 759.3 886.5 1013.8 1141 1268.2 1395.5 1522.7 1650 1777.2

Reaction time scores

Fig. 9.1. Normal Distributions Fitted by Mixture Analysis to TMT Data (top)
and RT Data (bottom). Number of Distributions Plotted Determined by
Significance Test.
Neuropsychological Tests 157

M = 2.4 SD = 1.9 Prevalence = 43%


M = 17.9 SD = 9.2 Prevalence = 57%
Estimated total distribution

8 -

nn
0.9 4.5 8.1 11.8 15.4 19 22.6 26.2 29.9 33.5 37.1

CORE scores

M = 27.6 SD = 8.0 Prevalence = 100%

8.9 12.4 15.9 19.5 23 26.5 30 33.5 37.1 40.6 44.1

SYMPTOM scores

M = 23.4 SD = 7.9 Prevalence = 100%

8.9 12.4 15.9 19.5 23 26.5 30 33.5 37.1 40.6 44.1

Hamilton Depression Rating Scale scores

Fig. 9.2. Normal Distributions Fitted by Mixture Analysis to CORE Scores


(top), SYMPTOM Scores (middle) and Hamilton Scores (bottom). Number of
Distributions Plotted Determined by Significance Test.
158 I. Hickie
ment is consistent with the neuropsychological phenomena encountered
in melancholia. Further, our MRI study demonstrates that structural
changes in the subcortical white matter are associated with impaired
reaction time in patients with melancholia. These associations support
the model that subcortical pathophysiologies can result in syndromes
characterised by mood disturbance, psychomotor change and a specific
pattern of neuropsychological impairment.
The neuropsychological data refute a number of potential key criti-
cisms of the CORE system. First, that CORE score ratings, and specifi-
cally the bimodal distribution of CORE scores, simply reflect a system-
atic rating bias among MDU team members. It could be argued that
clinicians working together in a research team, and particularly clini-
cians who share a common view of the nature of depressive sub-types
(and are testing a binary vs. unitary model of clinical typology), are
most likely to categorise patients as melancholic or non-melancholic and
then unwittingly produce CORE scores consistent with that categorical
preconception. As such, rater bias could account for the bimodal dis-
tribution of CORE scores. When the same bimodal distribution is noted
for independent measures of psychomotor speed, as was the case in the
MDU cohort, then it is more plausible to conclude that while overall de-
pression severity demonstrates a unimodal pattern, specific constructs
such as psychomotor speed (and probably psychotic phenomena) occur
categorically.
Secondly, psychomotor change and cognitive impairment are usually
assumed to be simply functions of increasing depression severity. From
this perspective, the detailed measurement of either provides few aetio-
logical or treatment insights beyond those provided by a global notion of
depression severity. Further, if all key depressive constructs behaved in a
similar dimensional fashion to global severity (that is, all demonstrated
comparable unimodal distributions), then this would provide support for
the notion that a common pathophysiological process underpins most de-
pressive disorders. The neuropsychological data, which are not subject
to the potential clinical bias of CORE scores, provide strong support for
a non-continuous distribution of at least some aspects of cognitive im-
pairment in depression. This pattern mirrors that of CORE scores, and
the contrast with the unimodal distribution of global severity ratings
(HRSD and SYMPTOM scores) strongly argues that not all depressions
are underpinned by a shared pathophysiological process (the so-called
final common [CNS] pathway). The non-continuous patterns of psy-
chomotor change and cognitive speed are more consistent with the view
Neuropsychological Tests 159
that at least one depressive sub-type (usually termed "melancholia") is
distinguished from other depressive disorders by observable psychomo-
tor change and a recordable neuropsychological deficit (largely of the
subcortical type).
10
Validity of the CORE Measure:
III. Outcome and Treatment Prediction
GORDON PARKER
IAN HICKIE
CATHERINE MASON

Introduction
As noted in Chapter 2, suggested defining characteristics of melancholia
have included selective response to physical treatments (i.e., antidepres-
sant drugs and electroconvulsive therapy) in addition to clinical features.
Again in that chapter, we noted support for the view of Nf Brolchain
(1979) that classification should be restricted "in the first instance, to
mental state items," before subsequently examining how aetiological and
other factors (including treatment) relate to the clinically defined classes.
In Chapter 6 we noted the reasons leading to revision of DSM-III melan-
cholia criteria, with Zimmerman, Black and Coryell (1989) pointing out
that the DSM-III criteria differentiated patients along a severity dimen-
sion, and that those criteria "did not predict treatment response."
If melancholia and non-melancholic depression have differential natu-
ralistic and/or treatment outcome patterns, and if the CORE measure is
a valid measure of melancholia, then CORE scores should show predictive
validity in outcome studies. Before reviewing several of our studies, we
first note some evidence supporting differential outcomes for melancholic
and non-melancholic depression.
The literature on naturalistic outcome is necessarily restricted by (i)
the variable validity of criterion measures of melancholia or endogenous
depression; (ii) intrinsic difficulties in measuring course of illness; (iii)
co-morbidity issues - notably the presence or absence of anxiety, per-
sonality disorder and medical illness; (iv) non-illness variables - such as
service access and sophistication; and by (v) varying definitions of the
term "naturalistic." As most long-term studies have involved depressed
subjects initially recruited during treatment, and who subsequently may
have had prophylactic interventions and/or subsequent episodes, we

160
Outcome and Treatment Prediction 161
know little about the true natural history of the depressive disorders,
or of any intrinsically differential pattern between types over time.
After conceding such caveats, there is, however, evidence that, in the
short term, melancholia has a distinctly lower rate of natural, sponta-
neous or placebo-induced remission. The review by Zimmerman et al.
(1989) provides evidence indicating that non-endogenous patients more
frequently respond to placebo. In addition, quantification of such a
differential effect has been provided by Fairchild et al. (1986), who es-
timated a placebo (or non-treatment) response of 6% in those with en-
dogenous depression compared to 54% in those with non-endogenous
depression.
Two long-term studies have provided some information on differential
patterns over time. Andrews et al. (1990a) reported a 15-year follow-up
of an Australian sample of depressed patients who had made a clear re-
covery from the index episode. Those with a baseline diagnosis of endog-
enous depression were more likely to have subsequently both remained
well and been hospitalised. The authors interpreted the pattern of en-
dogenous depression then as being recurrent and severe, but with suffer-
ers having neither a continuous illness nor significant inter-episode dis-
ability. By contrast, those with a baseline diagnosis of non-endogenous
depression were subsequently more likely to have prolonged periods of
minor symptomatic disturbance interspersed with franker episodes. In
the second study, Duggan, Lee and Murray (1991) reviewed patients ad-
mitted to the Maudsley Hospital, London, with a depressive illness some
18 years previously. They concluded that those with an endogenous de-
pressive disorder had a poorer global outcome.
In terms of active treatment, Zimmerman and Spitzer (1989) claimed
that "No study of antidepressant medication or ECT has supported
DSM-III'S suggestion that 'melancholies are particularly responsive to
somatic therapy."5 Nevertheless, after a review of 11 medication stud-
ies, they concluded that there was "a consistent trend toward superior
treatment response in the melancholic/endogenous group." Goodwin
and Jamison (1990) reviewed the literature on tricyclic antidepressants,
noting that "Relatively high response rates generally have been found
in patient groups predominantly characterized by so-called endogenous
symptoms. ... Conversely, relatively low response rates have been noted
among patients whose depressions are accompanied by overt delusions or
are characterized in the old terminology as neurotic or reactive." Rush
and Weissenburger (1995), in their review of melancholia for DSM-IV,
noted stronger support for melancholic sub-typing predicting a supe-
162 G. Parker, I. Hickie and C. Mason
rior response to antidepressants among depressed inpatients compared
with outpatients. They also recognized that pre-treatment retardation
had been identified as differentiating medication responders from non-
responders.
In relation to ECT efficacy, most reports (e.g., Abrams and Vedak
1991; Buchan et al. 1992) emphasise the predictive significance of melan-
cholic symptoms, psychotic features and/or the presence of psychomotor
disturbance. In their DSM-IV review of nine relevant papers, Rush and
Weissenburger (1995) observed that "Endogenous depressions fare bet-
ter with ECT than do non-endogenous depressions . . . [and that] . . .
there are no negative studies in this regard."
We conclude, then, that if the CORE measure provides a valid estimate
of melancholia, it should certainly predict response to ECT, probably
predict antidepressant medication response and possibly predict differ-
ential medium- and long-term outcome for those with differing CORE
scores. We now overview several of our studies considering the capacity
of the initial and final CORE measures to predict outcome.

A Study of Medium-term Outcome


As part of our CORE I study, we attempted to review progress of the
initial sample (N = 310), and were able to encourage 83 to attend
the MDU for a follow-up 20 weeks after initial assessment, and 75 to
attend a similar 1-year follow-up. We contrasted (Parker et al. 1992b)
those with melancholia (defined as meeting relevant DSM-ili, RDC and
CORE II criteria - as detailed in Chapter 6) with the remaining non-
melancholies. At each review, the research psychiatrists were required
to decide whether patients still met criteria for caseness.
By definition, all subjects met the caseness criterion at baseline, while,
at 20 and 52 weeks, 45% and then 29% of the melancholies as compared
to 39% and 26% of the non-melancholies met criteria. Severity (assessed
by Hamilton and Zung depression measures) was established as greater
for those with melancholic depression at baseline, while end point sever-
ity (at both 20 and 52 weeks) appeared similar for both melancholic and
non-melancholic depressive groups.
While we examined a number of predictors of improvement, our focus
here is on the CORE measure (here the CORE I version) as a predictor
of improvement, and of differential improvement across the melancholic
and non-melancholic groups. Remission by 20 weeks (i.e., no longer be-
ing a depressed case) was not predicted by baseline CORE scores when
Outcome and Treatment Prediction 163
examined in relation to the whole sample and within the non-melancholic
sub-sample. For patients, however, in the melancholic group, those in
remission at 20 weeks were somewhat more likely to have lower base-
line CORE scores (28.6 vs. 32.0, t = 2.0, n.s.), while those in remission
at 52 weeks were significantly more likely to have had lower baseline
CORE scores (27.9 vs. 35.4, t = 3.7, P < 0.001). Symptom reports
of psychomotor disturbance (i.e., both retardation and agitation) were
non-predictive.
It is interesting to consider these findings against those from the Lon-
don study of Duggan et al. (1991) having noted earlier their conclusion
that meeting criteria for endogenous sub-type predicted a poorer out-
come over 18 years. When those authors examined a set of individual
baseline clinical predictors, against their expectation that psychotic fea-
tures might drive the "predictive power," they established two others (re-
tarded activity and abnormal rate of speech) as actual predictors of poor
outcome, and concluded that "Retardation, a criterion of endogenous-
type depression, gives the continuum much of its predictive power."

Conclusions. While our naturalistic study has a number of limitations


(e.g., low compliance, assessment of improvement status being based
on cross-sectional time points), findings for the one-year follow-up are
compatible with one respected study in indicating that psychomotor
disturbance was a predictor of a somewhat poorer outcome. More im-
portantly, in relation to the objectives of this chapter, we established
that CORE scores assessed at baseline did have predictive potential - at
least for those assigned to a melancholic class.

CORE I Scores as a Predictor of Antidepressant Drug


Response
In developing the CORE I measure, we had the provisional rating sched-
ule completed by 30 volunteering Sydney psychiatrists for 300 of their
depressed inpatients and outpatients. Over the next 10 weeks the clin-
icians were requested to document details on any antidepressant medi-
cation prescribed and rate the level of clinical improvement, while the
patients completed Zung self-report and visual analogue depression mea-
sures.
We then assessed (see Parker and Hadzi-Pavlovic 1993) whether base-
line CORE scores predicted response to antidepressants in the 37 patients
so treated for at least three weeks and for whom details were available
164 G. Parker, I. Hickie and C. Mason

about their clinical status six weeks after initial assessment. We exam-
ined CORE scores both categorically and dimensionally as predictors.
The categorical analysis involved comparison of those scoring 25 or
more on the CORE I measure (HIGH CORE or putative melancholia)
versus residual subjects (LOW CORE or putative non-melancholia). In
this naturalistic study, we observed that HIGH CORE subjects were more
likely to have received antidepressants (78% vs. 57%) and, for those
who received antidepressants, 72% of the HIGH CORE and 37% of the
LOW CORE were judged as clinically improved at the six weeks' assess-
ment. While this finding supported the hypothesis, the naturalistic de-
sign meant that two variables (HIGH CORE status, and receipt of an-
tidepressant medication) might have confounded each other. That is,
as those in the HIGH CORE category were both more likely to improve
and to receive antidepressant medication, any differential efficacy of that
medication on melancholic and non-melancholic depression was difficult
to establish.
We also undertook several dimensional analyses. First, we used a set of
comparative multiple regression analyses examining the extent to which
three measures of melancholia predicted percentage reduction in Zung
self-report depression scores. Neither a DSM-ni diagnosis of melancholia
(F = 0.15, P = 0.70) nor a clinical diagnosis of endogenous depression
(F = 1.85, P = 0.18) was a significant predictor, but the CORE score
(F = 5.44, P < 0.025) did predict improvement. Next we undertook a
dimensional analysis essentially testing for a dose-response effect. Thus,
in our sample of depressed patients receiving antidepressants, we pursued
whether a progressively increasing CORE score predicted a progressively
better outcome. To fit the logit model, the CORE score (or dose) was di-
vided into five ranges (i.e., 16-21, 22-27, 28-33, 34-39, and 40-45), with
patients falling in any range receiving the range's mid-point CORE score.
Recovery (or response) was defined by reference to (in order) clinician
judgment of recovery, a set improvement in visual analogue scores or a
set improvement in Zung scores (with the three measures being used to
increase the sample size to 37). In this type of analysis, model fitting
provided, at each dose level, estimates of the proportion expected to
recover if there was a linear relationship between CORE scores and the
log of the odds of recovery.
Figure 10.1 plots the proportions of recovered patients expected un-
der the model and the proportions actually observed. Analyses detailed
in our report (Parker and Hadzi-Pavlovic 1993) provided modest evi-
Outcome and Treatment Prediction 165

§ -,

• Percentage observed recovering


9 Percentage expected to recover (CORE not in model)
A Percentage expected to recover (CORE in model)

16-21 22-27 28-33 34-39 40+

CORE score range (model used log of mid-point of range)

Fig. 10.1. Proportion of Patients Expected to Recover under the Logistic


Model, and Proportion Actually Recovering. Reprinted from Parker and
Hadzi-Pavlovic (1993).

dence that, for those receiving antidepressants, a higher CORE score was
associated with an increased probability of recovery.

Conclusions There were numerous limitations to this pilot study (e.g.,


it was open rather than controlled; the number of subjects was small;
types and doses of antidepressants as well as compliance details were not
assessed; our predictor measure was the CORE I system rather than the
166 G. Parker, I. Hickie and C. Mason

final CORE). Nevertheless, the CORE I system, whether examined as a


categorical measure of melancholic type or as a dimensional measure of
psychomotor disturbance (essentially retardation for the CORE I mea-
sure), was predictive of a better response to antidepressant medication.
While results support the predictive validity of the CORE measure, and
are compatible with the view that melancholia is preferentially respon-
sive to true benefit from antidepressant medication (as opposed to a
disguised natural remission or placebo response), controlled studies are
clearly required. Such studies would benefit from large sample sizes,
examination of varying antidepressant drugs, analyses considering other
factors influencing remission and, as suggested by Zimmerman et al.
(1989), pursuit of the treatment prediction that "melancholies are par-
ticularly unresponsive to placebo."

An Initial ECT Study


Following development of the CORE I measure, an open study was
undertaken of 35 depressed patients who had been consecutively recom-
mended for ECT (Hickie, Parsonage and Parker 1990). In addition to
examining the predictive validity of the CORE measure, we were keen
to compare the predictive capacities of psychomotor disturbance and
psychotic features, as both have been proposed as the strongest clinical
predictors of ECT response (Hobson 1953; Nystrom 1965; Carney et al.
1965; Weckowicz, Yonge and Cropley 1971; Kendell 1981; Johnstone,
Deakin and Lawler 1980; Brandon, Cowley and McDonald 1984).
Unfortunately for one of our objectives (i.e., to test the comparative
utility of ECT for melancholic and non-melancholic depression - at least
as diagnosed by the CORE system), all 35 patients had received a diag-
nosis of endogenous or psychotic depression from their treating clinician.
Each patient was assessed independently by two raters - who completed
the CORE measure and the 21-item Hamilton scale (with retardation and
agitation items from the latter summed to form a composite psychomo-
tor disturbance item for comparison against CORE scores). Measures
were repeated after the ECT course (mean number of treatments being
10.5), while the rater made a judgement about overall clinical improve-
ment. Outcome was also assessed by calculating percentage improve-
ment in Zung and Hamilton depression scales.
The subjects (14 males, 21 females; mean age = 63 years; mean Hamil-
ton = 25.7; mean Zung = 60.2) had a mean baseline CORE score of 29.6,
with all but 3 having a CORE I score of 21 or more (the cut-off assigning
Outcome and Treatment Prediction 167
those to a melancholia class). Twelve had psychotic features (delusions
and/or hallucinations), and from clinical observation raters assigned 9
as principally agitated and 26 as principally retarded.
In the whole sample, the baseline CORE I score was the strongest pre-
dictor of global response to ECT (r = +0.28), when compared with the
Hamilton psychomotor items (r = +0.14), the total 21-item Hamilton
score (r = -fO.13) and the baseline Zung score (r = —0.03). Compared
to the baseline Hamilton, the CORE score was a superior predictor within
sub-groups (r = 0.59 vs. r = 0.30 for psychotic depressives; 0.45 vs. 0.20
for retarded depressives; and —0.15 vs. —0.12 for agitated depressives),
but of similar predictive strength to the psychomotor items from the
Hamilton.
Recognising that the results were somewhat counter-intuitive for the
agitated sub-sample (i.e., a trend for higher scores on the CORE I
measure to be associated with less improvement), we then focussed our
multivariate analyes on the residual 24 retarded subjects. A stepwise
multiple regression analysis entered three pre-treatment clinical predic-
tors - (i) CORE score, (ii) Hamilton score (as a measure of severity) and
(iii) presence of psychotic features - with the outcome variable being
percentage improvement in Hamilton scores. In this analysis, the CORE
score emerged as the only significant predictor (r = 0.45, P = 0.03),
explaining 20% of the variance.

Conclusions As our sample did not contain non-melancholic depres-


sives, we were unable to undertake categorical analyses contrasting HIGH
CORE (putative melancholia) and LOW CORE (putative non-melancholia)
subjects, to test the accepted wisdom that ECT is more effective for
melancholic depression. We nevertheless examined the CORE measure as
a dimensional predictor of ECT improvement and demonstrated that,
while it had slight predictability in the whole sample (r = 0.28), it
was a modest predictor for those with a retarded clinical presentation
(r = 0.45) and quite a strong predictor in those with psychotic depres-
sion (r = 0.59) - presumably reflecting higher CORE scores, as the pres-
ence of psychotic features itself was not a strong predictor of outcome
(alone or in the multivariate analysis). The CORE'S lack of prediction
for those with an agitated clinical presentation (r = —0.15) most likely
reflected the lack of agitation items in the CORE I measure, and with re-
sults for this sub-group weakening the overall predictability of the CORE
measure across the whole sample. In our multivariate analysis we sought
to redress the overlap and shared domains occupied by three measures
168 G. Parker, I. Hickie and C. Mason
(i.e., CORE, Hamilton and psychotic features). The CORE score emerged
as the only significant predictor in that analysis, suggesting that, af-
ter accounting for any shared variance, those two other variables had
no residual components predicting ECT improvement. Results were en-
couraging, then, in indicating that the CORE I measure had predictive
validity in an open study of patients referred for ECT.

The Final CORE Measure as a Predictor of ECT Response


After developing the final CORE measure, with its motoric agitation as
well as retardation items, we examined its capacity to predict ECT out-
come (Hickie, Mason, Parker and Brodaty, in preparation) - examining
both total CORE and CORE sub-scale scores. Again, we compared the
CORE'S predictive capacity against (i) the agitation and retardation psy-
chomotor items from the 21-item Hamilton scale, (ii) presence of psych-
otic features and (iii) severity of mood disturbance (as measured by the
full Hamilton scale score).
Patients were recruited from eight hospitals in the Sydney metropoli-
tan area. Following exclusion of those with a primary organic disorder
or atypical or schizophrenic psychoses, the final sample comprised 81
subjects with a primary depression. After a detailed baseline clinical in-
terview, CORE and Hamilton ratings were made by two psychiatrists (IH
and CM). After the ECT course was completed (its duration determined
by the treating clinician and with a mean of 11.3 treatments), measures
were repeated and clinical outcome assessed. Improvement was quanti-
fied by effect size changes in (i) Hamilton and (ii) the DSM-III-R Global
Assessment of Function (GAF) Scale, as well as by clinical improvement
(assessed by a five-point global rating).
The sample had a mean age of 67 years, a female preponderance (1.9:1)
and mean depressive episode duration of 26 weeks. Mean Hamilton
and CORE scores were 28.8 and 24.8 respectively. Psychotic symptoms
were present in 57%. In terms of melancholic sub-typing, 79% scored as
having endogenous depression on the Newcastle score (i.e., a score of 6 or
more), 95% scored 8 or more on the CORE score (putative melancholia),
and only 5% did not receive an MDU clinical diagnosis (see Chapter 6)
of endogenous or psychotic depression.
Table 10.1 summarises the capacity of the following baseline measures
to predict improvement on the several outcome measures:

(i) total CORE score;


Outcome and Treatment Prediction 169

Table 10.1. Capacity of baseline CORE score and its sub-scales,


as well as Hamilton and psychoticism variables, to predict
improvement on three parameters

Global
Effect size units outcome
Predictor Hamilton GAFf scale
Total CORE score 0.42*** -0.38*** 0.27*
CORE sub-scale
Retardation 0.33** -0.30** 0.23*
Agitation 0.23* -0.24* 0.13
Non-interactiveness 0.40*** -0.33** 0.24*

Total Hamilton — -0.24* 0.17


Hamilton items
Retardation 0.21*** -0.19** 0.08*
Agitation 0.18 -0.17 0.13
Psychotic symptoms 0.44 -0.34 0.27
f Global assessment of functioning from IDSM-III-R.
**P < 0.001; **P < 0.01; * P < 0.05.

(ii) CORE sub-scale scores;


(iii) the Hamilton depression scale psychomotor disturbance items;
(iv) total Hamilton score (i.e., depression severity); and
(v) the presence of psychotic symptoms.
Correlation coefficients summarised in Table 10.1 suggest that the best
(and rather equivalently successful) predictors of improvement were the
total CORE score and the presence of psychotic symptoms, and that
these were superior to the total Hamilton score. Each of the CORE
sub-scale scores was a significant predictor of improvement, with CORE
non-interactiveness scores being slightly superior to the CORE motoric
scale scores.
As univariate analyses had demonstrated inter-relationships between
these several predictors (i.e., CORE with total Hamilton = +0.54; CORE
with psychotic symptoms = +0.36; Hamilton with psychotic symptoms
= +0.46), we entered all three predictors in a series of stepwise multiple
regression analyses. As predictors of global outcome, only the presence
of psychotic features entered the final equation (/? = 0.27, P < 0.05);
while as predictors of change in GAF score, only the total CORE score
170 G. Parker, I. Hickie and C. Mason

entered the final equation (/? = 0.38, P < 0.05). We further explored the
relative predictive capacities of those two features, undertaking analyses
of variance to determine whether the predictive capacity of psychotic
features could be accounted for by CORE-rated psychomotor disturbance,
and established that both variables made significant and independent
contributions. In an additional analysis, we sub-grouped subjects on
the basis of their being above or below the median CORE score (in this
population, 25), and then established four groups differing on the basis of
(i) being above or below the CORE median, and (ii) presence or absence of
psychotic features. The combination of high CORE scores and psychotic
features was distinctly associated with the best outcome on all three
measures.

Conclusions In this more definitive study (in testing the final CORE
measure in a large sample), we established the CORE measure as a strong
predictor of ECT response. The predictive validity of the CORE was es-
tablished as being independent of both psychotic features and of sever-
ity (presuming that the Hamilton scale is a measure of severity). In
comparison to the earlier ECT study, we clarified that one aspect of
psychomotor disturbance - agitation - was also predictive, along with
the other two domains (retardation and non-interactiveness) measured
by CORE scales.

Discussion
The CORE I measure was shown to have the capacity to predict one-
year outcome, and both antidepressant medication and ECT response,
results which further encouraged our development of the final CORE mea-
sure. As yet, only one treatment outcome study has been undertaken
with the final CORE measure, and that ECT study (overviewed here) has
also demonstrated a clear-cut predictive capacity.
If melancholia is associated with a superior response to antidepres-
sant medication and to ECT, and if the CORE system is a valid measure
of melancholia, the CORE system should have the capacity to predict
response to such physical treatments. Our positive findings therefore
support, but do not necessarily establish, that the CORE system is actu-
ally a valid measure of melancholia.
Both our antidepressant and ECT studies were limited by having very
high representations of melancholic patients. Ideally, the capacity for the
Outcome and Treatment Prediction 171
CORE system to predict treatment outcome would be tested in CORE-
assigned putative melancholic and non-melancholic classes (i.e., those
scoring less than 8 were contrasted with comparable numbers of those
scoring 8 or more), and with differential improvement across the classes
being tested. Our studies, involving few non-melancholic depressives,
risked the possibility that CORE scores would be non-predictive - for, if
melancholia responds to ECT and antidepressant medication, all of our
melancholic subjects would have been expected to improve, and differ-
ential or significant effects may not have been demonstrated. The lack of
non-melancholic subjects, particularly in the ECT studies, required us
to undertake dimensional analyses, with increasing CORE scores repre-
senting increasing levels of psychomotor disturbance. Our dose-response
analyses were consistent in suggesting that treatment (most strongly in
relation to ECT) was most likely to be effective at higher levels of CORE-
defined psychomotor disturbance. Such dose-response effects (akin to
that demonstrated for the DST in Chapter 8, where increasing CORE
scores were associated with increasing DST non-suppression rates) sug-
gest that both aetiological and treatment outcome research should pur-
sue perturbed domains (like psychomotor disturbance) dimensionally
and not only by contrasting the influence of their presence or absence.
11
Phenotypic Expression of Melancholia
Contrasted for Those with Bipolar and
Unipolar Illness Courses
PHILIP MITCHELL
AYSE SENGOZ

Introduction
Since Leonhard (1979) originally proposed the distinction between bipo-
lar and monopolar (unipolar) forms of affective disorder, there have been
many attempts to validate this separation (Angst 1966; Perris 1966;
Perris 1990; Perris 1992). In particular, there has been considerable
interest in identifying pathophysiological or clinical features which may
differentiate the depressed phase of bipolar disorder from the depressed
phase of unipolar depression. Any such demonstration of distinguishing
features would be of considerable clinical value, particularly for first-
onset depressive episodes or when no accurate history is available. More
importantly, differing clinical presentations (or phenotypic expressions)
might indicate differential biological determinants.
Whilst a number of biochemical and endocrine investigations have
been undertaken, with some studies suggesting possible biological dif-
ferences between these syndromes (e.g., Roy et al. 1985; Carroll 1994),
there have been relatively few comparative studies of clinical features.
In this chapter we examine for differences in clinical features - and par-
ticularly for differences in psychomotor disturbance - between bipolar
and unipolar melancholic depressed patients, as psychomotor retarda-
tion has traditionally been considered to be more prevalent in bipolar
depression (Goodwin and Jamison 1990).
In relation to clinical features, Leonhard (1979) emphasised both vari-
ability between episodes and the polymorphic nature of both the de-
pressed and manic phases of bipolar disorder, with the term "polymor-
phic" used to describe his observation of "signs of lability toward the
other pole." Leonhard also believed that "partial states" (i.e., those
with the absence of essential single symptoms, for example, "unpro-

172
Phenotypic Expression in Bipolar and Unipolar Illness 173
ductive mania") were characteristic of both phases of the bipolar form,
in comparison to the "pure" presentations of unipolar melancholia or
unipolar mania.
The original validation study of Leonhard's classification by Perris
(1966) did not demonstrate differentiation of clinical features between
unipolar depression and the depressed phase of bipolar disorder, though
phenomenological examination was not extensive. A number of subse-
quent investigations (see tabulated review in Mitchell et al. 1992) have
suggested differences in clinical features, but these have been incon-
sistent. Many differences and inconsistencies are likely to have been
generated by differing definitions of "unipolar." To many, unipolar de-
pression is a residual and heterogeneous class comprising all non-bipolar
depressed patients. In this chapter, we preserve Leonhard's distinction -
limiting, in effect, unipolar depression to non-bipolar melancholic depres-
sion. Despite inconsistencies across previous reports, a picture emerges
of unipolar depression being characterised by both longer duration and
greater severity of episodes and for the following features to be over-
represented: unvarying quality, suicidal intent, initial insomnia, weight
loss, muddled thoughts and somatic complaints. In contrast, bipolar de-
pression has been held to be associated with more severe diurnal mood
variation (morning worsening), lability of mood, hypersomnia and dere-
alisation. Psychotic features have been described as over-represented in
both bipolar and unipolar depression.
Psychomotor disturbance in bipolar and unipolar depressed patients
has been assessed by both observational clinical measurement and for-
mal actimetric methods. Most, but not all, investigations which have
used formal continuous actimetric monitoring techniques have reported
higher psychomotor activity levels in unipolar patients. Thus, Beigel
and Murphy (1971) and Kupfer et al. (1974) found higher activity lev-
els in the unipolar depressed patients, while Kuhs and Reschke (1992)
found no significant differences. Kupfer et al. (1974), however, did not
match patients for age and sex, unlike the other two studies.
A number of other studies (Perris 1966; Abrams and Taylor 1974;
Dunner, Dwyer and Fieve 1976; Katz et al. 1982; Popescu et al. 1991)
have reported on clinically assessed psychomotor disturbance, examin-
ing for both agitation and retardation, with findings being discrepant.
Instruments employed have included scales such as the Salpetriere Rat-
ing Scale of Widlocher (1983), which was employed by Popescu et al.
(1991). Despite a lack of consistency, most clinical descriptions (e.g.,
174 P. Mitchell and A. Sengoz

Goodwin and Jamison 1990) continue to describe psychomotor retarda-


tion as being more prevalent in bipolar depressed patients.
While longitudinal course is not the focus of this chapter, differences in
long-term patterns between bipolar disorder and recurrent unipolar de-
pression have also been demonstrated. Recently, for example, Winokur
et al. (1993) reported chronicity to be more prevalent in unipolar pa-
tients, with multiple episodes being more common in those with bipolar
disorder. In this chapter, we overview analyses undertaken in the two
independent CORE I and CORE II samples.

The CORE I Study


Methods
As described in Chapter 6 and detailed in a separate report (Mitchell
et al. 1992), we rated an extensive list of symptoms and mental state
signs for the nadir of the present episode. Clinical episode diagnoses, cor-
responding broadly to ICD-9 categories, were made from data obtained
at the semi-structured interview, and reviewed during treatment and
at follow-up, while DSM-III and RDC diagnoses were made from features
elicited at baseline.
As we sought a sub-sample of (unipolar and bipolar) melancholic pa-
tients, and as our findings would be driven by the rigour with which
melancholia was refined, we included only those depressed patients who
met each of the DSM-III, RDC and CORE criteria for melancholia (or en-
dogenous depression). These diagnostic criteria resulted in 138 of the
305 patients being assigned as composite melancholies. Bipolar melan-
cholic patients were then defined as those fulfilling RDC criteria for past
manic or hypomanic episodes, with the clinical data sought from both
the initial assessment interview and previous medical records, and with
assignment to the bipolar category being made only if supportive fea-
tures were recorded as being unequivocally present.
Of the initial 138 composite melancholies, 27 were rated as bipolar (17
manic, 10 hypomanic). These 27 patients were then matched for age and
sex with 27 of the remaining 111 unipolar composite melancholies, each
bipolar patient being paired with a unipolar patient of the same sex and
of the closest age. Comparisons involved the two-tailed paired £-test for
dimensional data and, for categorical data, the McNemar test.
Phenotypic Expression in Bipolar and Unipolar Illness 175
Results
Sample Characteristics. The sample comprised 17 females and 10
males in the matched groups, now termed the bipolar and unipolar
groups. The mean ages (SD) of these respective groups were 47.4 (14.5)
and 47.2 (14.4) years, while age of onset of the first episode of depression
was also similar (32.5 vs. 35.0 years).
The bipolar patients had significantly briefer current episodes than
the unipolar (28 vs. 70 weeks) and were more likely to have experienced
a previous depressive episode (92% vs. 60%). The bipolar and unipo-
lar had similar Zung self-report depression scores (54.5 vs. 55.3) and
clinician-rated Hamilton depression scores (21.9 vs. 20.9). Similar per-
centages (15% vs. 22%) had received a clinical diagnosis of psychotic
depression for the episode.

Mental State Signs. The two groups were compared on the 30 mental
state signs listed in Table 6.2 (Chapter 6). The bipolar melancholies were
significantly (P < 0.05) more likely to rate as nihilistic and less likely
to rate as having slowed movements. There was, however, a general
trend for bipolar patients to be rated as less retarded on several other
retardation items (i.e., slumped posture; immobility; muteness; retarded
movement) and to be rated as more agitated (i.e., agitated movements;
querulous and irritable). The decreased frequency of psychomotor re-
tardation in the bipolar group appeared unrelated to concurrent drug
treatment, as identical percentages (4%) of subjects in each group were
taking phenothiazines. On the derived total CORE I score the bipolar
subjects returned scores similar to those of the unipolar (29.4 vs. 30.2)
subjects.

Depressive Symptoms. The two groups were compared on 37 symp-


toms, most being listed in Table 6.1 (Chapter 6). There were no sig-
nificant differences in symptom ratings between the groups, although
bipolar patients tended to be more likely to complain of subjective agi-
tation, and to demonstrate more vegetative abnormalities such as early
morning wakening, mood and energy worse in the morning and greater
loss of appetite.
176 P. Mitchell and A. Sengoz
The CORE II Study
Methods
The patients selected for this study were derived from an extension of
the CORE II sample, and comprised 427 consecutively diagnosed pri-
mary depressive patients. The methodology was largely similar to the
earlier unipolar versus bipolar study, with the key difference being that
only DSM-III-R diagnoses for melancholia were used. In the first study,
allocation to the composite melancholic group required meeting relevant
RDC, DSM-m and CORE criteria. The last criterion, however, risked a
bias emerging from comparison of psychomotor disturbance in groups
already defined (at least partially) by that very criterion.
As in the first study, our second-order decision about bipolar sta-
tus was assessed with reference to RDC criteria for past manic or hy-
pomanic episodes. This resulted in 25 DSM-III-R melancholic subjects
being rated as bipolar (20 manic, 5 hypomanic). Each bipolar melan-
cholic was matched with a unipolar melancholic patient of the same sex
and of the closest age.

Results
Sample Characteristics. The sample comprised 16 females and 9
males in the matched groups, with mean ages of 46.8 (bipolar) and
46.6 (unipolar) years. There were no significant differences in either the
age of onset of the first episode of depression (30.3 vs. 33.3 years) or the
duration of the current episode of depression (37.4 vs. 67.5 weeks), al-
though there was a clear trend for the episode to be briefer in the bipolar.
The bipolar patients also tended to have more previous episodes (6.8 vs.
3.4). In terms of depression severity for the current episode, no signifi-
cant differences were found for the Zung or Hamilton-17 or on a Global
Assessment of Functioning measure. The bipolar group did, however,
have higher Newcastle scores than the unipolar matched group (6.9 vs.
4.1). While there were no significant differences between the proportions
of those who had experienced hallucinations (12% vs. 8%) or delusions
(24% vs. 24%) during the current episode, there was a trend for more
bipolar than unipolar patients to have been psychotic during a previous
episode (52% vs. 20%).

Mental State Signs. There was no significant difference in the total


CORE score between the two groups (bipolar 12.6 vs. unipolar 10.0).
Phenotypic Expression in Bipolar and Unipolar Illness 111
On examination of specific items, there were, as in the previous study,
strikingly few suggested differences; the only formal difference was that
bipolar melancholies were more likely to demonstrate shortened verbal
responses (52% vs. 20%).

Depressive Symptoms. The two groups were contrasted on some 46


symptoms, including all those listed in Table 6.4 (Chapter 6). Only
three significant differences were established. The bipolar patients were
less likely to report initial insomnia (24% vs. 56%) and suicidal thoughts,
plans or attempts (48% vs. 80%), and were more likely to report sleeping
longer than usual (28% vs. 0%).

Discussion
Apart from the 1971 study of Beigel and Murphy (1971), our two studies
are the only ones that have limited comparison of bipolar and unipo-
lar depressed patients to strictly defined groups of melancholic subjects
closely matched for age and sex. Additionally, few previous reports have
utilised operational criteria, particularly for past manic or hypomanic
episodes.
Replication across the two studies of two clinical course variables (i.e.,
bipolar melancholies having briefer episodes at assessment, and more
previous episodes) suggests the likelihood of real differences in the lon-
gitudinal pattern of illness for these two disorders.
The main overall impression from these studies, however, is the re-
markable similarity between bipolar and unipolar patients in terms of
cross-sectional clinical features, especially when one would expect the
multiple tests to produce a number of Type I errors. Total CORE scores
were not significantly different between the bipolar and unipolar groups
in both studies, indicating a similar level of psychomotor disturbance.
Item analyses in the CORE I study suggested that slowed movements
were more frequent in the unipolar, whereas in the CORE II study we
quantified the length of verbal responses as shorter in the bipolar pa-
tients. The possibility that these differences may represent chance or
type I errors is suggested by the rarity of significant differences across
the two studies for the signs (2/49) and the lack of consistency in posi-
tive findings across the two studies. The few differences in the CORE II
study are of particular interest, as the bipolar melancholic patients had
higher Newcastle scores, suggestive of a greater likelihood of endogeneity
features.
178 P. Mitchell and A. Sengoz

Although it is difficult to reconcile our findings with those of the


methodologically strict study of Beigel and Murphy (1971), the differ-
ences with the other positive studies (Kupfer et al. 1974; Katz et al.
1982) may be explained by their lack of close matching. Our findings
are, however, consistent with the more common finding of no differ-
ences in rates of psychomotor disturbance between bipolar and unipolar
depressives (Perris 1966; Abrams and Taylor 1974; Katz et al. 1982;
Popescu et al. 1991; Kuhs and Reschke 1992).
Our results are also consistent with the majority of neuropsychologi-
cal assessments of psychomotor slowing. Using a number of standardised
tasks, most cognitive investigations have found no differences between
bipolar and unipolar depressed patients (e.g., Donnelly et al. 1982; Hoff-
man, Gonze and Mendlewicz 1985; Gunther et al. 1988). The only excep-
tion was Blackburn (1975), who found slower movement in the bipolar
depressives.
Turning to symptoms, there were no significant differences in the
CORE I study, while in the CORE II study, the bipolar patients
more frequently complained of hypersomnia, whereas unipolar patients
reported insomnia. These two findings have been reported previously
(Detre et al. 1972; Brockington et al. 1982).
In conclusion, these studies demonstrated few clinical differences be-
tween the bipolar and unipolar melancholic groups. At the least, our
findings suggest that the widely held belief that bipolar depressed pa-
tients are more likely to be psychomotor retarded than those with unipo-
lar melancholia (e.g., Goodwin and Jamison 1990) is difficult to substan-
tiate when psychomotor disturbance is assessed behaviourally by trained
clinicians. The overall similarity of cross-sectional clinical features sug-
gests that the phenotypic expression of melancholia is identical in those
who differ in terms of having bipolar or unipolar illness courses. Such
a close phenotypic resemblance argues for a commonality of biological
dysfunction at some level, despite the distinct genetic differences held
to exist between these two disorders.
12
Psychotic Depression: Clinical Definition,
Status and the Relevance of Psychomotor
Disturbance to Its Definition
GORDON PARKER
IAN HICKIE
DUSAN HADZI-PAVLOVIC

Introduction
We will first provide evidence of quite contrasting views about the sub-
typing of psychotic (delusional) depression. Next we will report two
studies that consider whether it has any specific or over-represented
clinical features (examining psychomotor disturbance in particular), and
then we will consider its status as a separate entity or a sub-type of some
other broader depressive condition. Our two studies were undertaken as
components of the CORE I and CORE II studies (reported in Chapter
6); methodological issues are only summarised here.

Psychotic Depression: Synonymous with, or Distinct from,


Melancholia? It is relatively easy to establish quite varying theoreti-
cal views about the relationship of psychotic depression to melancholia.
An early separatist (in viewing the condition as distinct) was Mauds-
ley (1895), who distinguished between "melancholia" and "melancholia
with delusions." More recently, Minter and Mandel (1979) suggested
that, in the United States, psychotic depression had "come to mean the
same as 'endogenous depression' or to mean a depression of severe pro-
portions, usually with endogenous symptomatology, not necessarily with
symptoms of psychosis." As well, psychotic and endogenous depression
have been used as synonymous terms by proponents of the contrasting
unitary (e.g., Kendell 1976) and binary (e.g., Kiloh et al. 1971) views of
depression classification. Thus, we observe two contrasting views - one
essentially regarding the condition as synonymous with melancholia, the
other regarding it as a separate disorder.

179
180 G. Parker, I. Hickie and D. Hadzi-Pavlovic
Classificatory Status
In most official classifications of psychotic depression, psychotic features
figure prominently, together with two variable constructs - functional
impairment and psychomotor disturbance. For instance, DSM-III (Amer-
ican Psychiatric Association 1980) criteria essentially required
(i) major depression to be present; and accompanied by
(ii) "gross impairment in reality testing" (evidenced by delusions, hal-
lucinations or depressive stupor).
The RDC system (Spitzer, Endicott and Robins 1980) allowed two sub-
categories
(i) "psychotic major depressive disorder" (either "probable" or "def-
inite") for those with major depression and having delusions or
hallucinations; and
(ii) "incapacitating major depressive disorder" (either "probable" or
"definite") to describe a psychotic depressive category essentially
on the basis of extreme severity or functional impairment.
The ICD-9 system (World Health Organization 1977) required, for an
affective psychosis, a severe disturbance of mood, accompanied by one
or more of the following: "delusions, perplexity, disturbed attitude to
self, disorder of perception and behaviour." In the "manic-depressive
psychosis, depressed type" sub-group, alternatives of "reduced activity"
and "restlessness and agitation" were described. The ICD-9 system also
allowed for a "reactive depressive psychosis," which was clinically similar
except for less diurnal variation in symptoms, association with a "serious
disturbance of behaviour" such as a suicide attempt, and provocation
by a major life stressor.
The DSM-III-R (American Psychiatric Association 1987) system distin-
guished between severity and psychosis in sub-classifying major depres-
sive episodes, and deleted the DSM-III option, allowing the capacity to
reach diagnosis on the basis of "depressive stupor." The DSM-III-R sys-
tem included a "severe, without psychotic features" sub-group, as well
as sub-groups with either mood-congruent or mood-incongruent psych-
otic features, with the latter groups defined entirely by the existence of
delusions and/or hallucinations. The DSM-IV (American Psychiatric As-
sociation 1994) system applies a "severe" specifier to those episodes of
major depression with psychotic features, with the psychotic features
being delusions or hallucinations (whether mood-congruent or mood-
incongruent). The ICD-10 system (World Health Organization 1992)
Psychotic Depression 181
similarly requires the depressive episode to be "severe," and the inclu-
sion of either delusions/hallucinations or depressive stupor to achieve
the diagnosis. It is clear then that official classificatory systems have
(i) consistently viewed psychotic depression as a severe condition, irre-
spective of its classificatory status;
(ii) variably required delusions and/or hallucinations as mandatory fea-
tures; and
(iii) variably allowed this diagnosis for depressed subjects with severe
psychomotor disturbance (e.g., depressive stupor) alone.
Varying views about the "stem" disorder are also evident. Thus,
Maudsley's distinction between melancholia with and without delusions
imputes melancholia as the base diagnosis. A typology stemming from
psychomotor disorder is apparent in Kraepelin's (1907) classification of
"depressive states" where he distinguished between those with (i) "sim-
ple retardation unaccompanied by any hallucinations or delusions" and
(ii) a "delusional form," characterised by the "presence of varied depre-
ciatory delusions, especially of self-accusation and of a hypochondriacal
nature, in addition to the evidence of retardation." Finally, the last
three DSM classifications have, in essence, superimposed features on a
base diagnosis of major depression, resulting in most published studies
seeking to clarify either the status or the clinical features of psychotic
depression using major depression rather than melancholia as the di-
agnostic comparison group. Such a comparison strategy clearly risks
confounding severity-based features with diagnosis-specific ones.

Illness and Clinical Features of Psychotic Depression


Identified in Empirical Studies
Two recent papers (Parker et al. 1991b; Schatzberg and Rothschild 1992)
have reviewed the published studies. Our review failed to suggest any
firm epidemiological differences between psychotic and comparison de-
pressives for age, age of initial onset, sex ratio, number of previous
episodes or morbid risk of major depression in relatives, apart from
the psychotic depressives who had an increased rate of bipolar disorder
amongst their relatives. The second review found strong evidence of ill-
ness consistency; those having psychotic depression were more likely to
have had not only delusional episodes in the past but the same delusional
themes repeated across episodes. In terms of clinical features, and with
most studies having used major depression (as opposed to melancholia)
182 G. Parker, I. Hickie and D. Hadzi-Pavlovic

as the comparison group, only a short list was generated in addition to


delusions and hallucinations in the two reviews. The most consistently
suggested additional feature was the presence or greater severity of psy-
chomotor change (agitation and/or retardation). In addition, there is
evidence of a more severe mood disturbance (with self-reproach and guilt
appearing to be the most qualitatively distinct mood symptoms).

Is Psychotic Depression a Separate Entity?


The view that psychotic depression may be a separate entity has been
argued less on the basis of any distinct clinical (especially psychotic)
features and more commonly on the basis of the condition showing a
poor response to antidepressants alone, as well as by a pattern of simi-
lar episodes over time (Nelson, Khan and Orr Jr. 1984; Glassman and
Roose 1981; Charney and Nelson 1981). Support for differential treat-
ment responses emerged from summary statistics for drug treatments
provided by Spiker et al. (1985) in the mid-eighties - which suggested
that one-third of delusional (psychotic) depressives respond to a tricyclic
antidepressant alone; one-half to an antipsychotic alone, and four-fifths
to combination antidepressant/antipsychotic drug therapy. Those au-
thors noted that the efficacy of ECT in psychotic depression appears
comparable to combination drug therapy. We subsequently undertook
a meta-analysis of a somewhat larger number of treatment studies (i.e.,
44), and reported (Parker et al. 1992c) that ECT appeared the most
effective treatment (with a mean effect size of 2.30), followed by com-
bination tricyclic/antipsychotic (ES = 1.56) and then by tricyclic alone
(ES = 1.16). In essence, our review differed from the Spiker et al. (1985)
overview in suggesting that
(i) tricyclic alone was probably not as ineffective as previously held;
(ii) combination tricyclic/antipsychotic was not so clearly superior to
tricyclic alone as previously held; while
(iii) our data indicated that bilateral ECT was quite superior to unilat-
eral ECT for the condition.
To date, the view that psychotic depression is a separate entity has
therefore been argued almost entirely on the basis of a differential treat-
ment response. However, clinical differences (increased psychomotor dis-
turbance and a more severe mood disturbance) and the suggestion of
some biological marker differences (e.g., increased HPA activity distur-
bance) could equally reflect nothing other than a more severe expression
Psychotic Depression 183
of depression (or of melancholic depression). If it is true that psychotic
and non-psychotic depressive disorders show distinctly differential treat-
ment responses, phenomenological definition of psychotic depression is
clearly of key importance. If, as has been conceded in some classificatory
systems, the condition may exist without delusions or hallucinations be-
ing present (or elicited), there is a need to define other key descriptors
(be they proxies for, or independent of, psychotic features) to allow its
clinical diagnosis. We therefore focus, both in this chapter and through-
out this book, on the relevance of psychomotor disturbance (along with
several other clinical features) to determining the definition and status of
psychotic depression, now reporting relevant material from our CORE I
and CORE II studies.

The CORE I Study


Introduction. As we sought to determine whether psychotic depression
might be a separate type of depression (as opposed to a more severe ex-
pression of depression, and in particular, melancholia) we examined for
specific and over-represented clinical features. As our clinical experi-
ence with such subjects suggests that they tend to be older, to be more
treatment-resistant and to have more chronic illness courses, we wished
to study such issues in a longitudinal study design. Respecting the long-
standing debate as to whether psychotic depression is synonymous with,
or a more severe expression of, melancholia, we formed a contrast group
of patients with non-psychotic melancholic depression.

Subjects and Assessment Procedure. The psychotic depressed sub-


jects were derived from the first 310 consecutively diagnosed primary
depressive patients assessed as inpatients or outpatients of the MDU,
with study details (including rating scales and the interview schedule)
described in a journal report (Brodaty et al. 1987) and in Chapter 6. Di-
agnostic information was generated from baseline assessment and from
historical information, as well as during treatment and at regular follow-
up reviews, so that clinical diagnoses could be refined or changed over
time. We had some patients whose delusions and/or hallucinations had
remitted by baseline assessment and, conversely, some who developed
such features after that assessment. Thus, in this study we provide
episode clinical diagnoses.
184 G. Parker, I. Hickie and D. Hadzi-Pavlovic
Defining Patient Groups. Three sub-groups of subjects were selected
for top-down analyses (i.e., preserving initial diagnostic allocations):

(i) Psychotic depressives (PDs), who met


(a) DSM-III criteria for melancholia and psychosis, and
(b) RDC criteria for definite endogenous depression and psychosis,
and
(c) our MDU clinical criteria for psychotic depression, which essen-
tially meant clinically evident delusions and/or hallucinations;
and where a subject was accepted as psychotic only if there was
unequivocal evidence of delusions or hallucinations.
(ii) Endogenous depressives (EDs), who met
(a) DSM-III criteria for melancholia, and
(b) RDC criteria for definite endogenous depression but who were
not psychotic, and
(c) our MDU clinical criteria for endogenous depression (defined
by the presence of a number of endogeneity symptoms), but
without evident delusions or hallucinations.
(iii) A sub-set of the EDs (termed MEDs) who were age- and sex-
matched with the PDs.

Results
One hundred and thirty-seven subjects (44% of the total sample) met
criteria for DSM-III melancholia, RDC definite endogenous depression and
clinical endogenous depression. Thirty-five (26%) were both psychotic
and endogenous (the PDs), while 102 (74%; the EDs) were endogen-
ous but not psychotic. Thus, 11% (35/310) of the overall sample were
diagnosed as having psychotic depression.

Sociodemographic, Historical and Illness Course. We summarise


analyses from our earlier report (Parker et al. 1991b). While there was
a similar female preponderance across the PDs and EDs (66% vs. 64%),
the PDs were more than a decade older than the EDs both at intake and
initial onset of any depressive episode. Although 91% of the PDs and
64% of the EDs were more than 40 years old at initial onset, only 14% of
the PDs, compared to 40% of the EDs, had their first episode commence
before the age of 30 years. Thus, the older age of the PDs would appear
Psychotic Depression 185
to reflect the comparative rarity of any episodes of depression (psychotic
or other) before 30 years of age in that group. The groups did not dif-
fer significantly in reporting a previous manic episode, or in reporting a
family history of depression or schizophrenia. The PDs were much more
likely than the EDs to have previously received combination antipsy-
chotic/antidepressant medication (55% vs. 17%) and/or ECT (66% vs.
31%). Duration of current episode (50 vs. 48 weeks) did not distinguish
the groups. The PDs and EDs did not differ in depression severity on
the self-report Zung scale, but the PDs did score higher on the clinician-
rated severity and endogeneity measures (56% higher on the Newcastle
Scale and 43% higher on the Hamilton). Apart from the matched age
variable, differences remained when the PDs were compared with the
MEDs.
Treatment for sample members was naturalistic and relied upon clini-
cal judgement. The majority of PDs (91%) were admitted to hospital for
the current episode, compared to two-thirds of the MEDs, but the PDs
had a much longer stay (88 vs. 39 days). The PDs were significantly
more likely to receive combination antidepressant/antipsychotic medi-
cation (77% vs. 24%) and/or ECT (66% vs. 30%) than the MEDs. We
reviewed the course of the PDs and MEDs for the 12 months after intake
assessment, allowing three broad outcomes. The clearest difference was
that the PDs were more likely to have the poorest outcome (48% vs.
26%), a finding matched by a higher hospital readmission rate (48% vs.
26%). Our one-year follow-up data showed a particularly poor course
for the PDs, with only 17% both recovering and remaining well over that
interval - a finding in apparent contrast with the literature which sug-
gests high rates of success for combination antipsychotic/antidepressant
medication and/or ECT. We offer three interpretations. First, our re-
ferral patterns resulted in more intractable and/or recurrent disorders.
Second, the claims for high treatment success may refer only to imme-
diate therapeutic response - when most of those studies failed to follow
samples over extended periods. Third, as our PDs were, on average, a
relatively old group, and as in some reports depression per se has a poor
prognosis in the elderly (e.g., Burvill, Stampfer and Hall 1986), the poor
illness course for our PDs may well reflect an age effect rather than the
nature of PD alone.

Clinical Features. We summarise analyses from an earlier report (Par-


ker et al. 1991a), first examining signs and then symptoms.
186 G. Parker, I. Hickie and D. Hadzi-Pavlovic

Table 12.1. Presence and severity of the CORE I signs in the PDs
and MEDs

Presence Odds
CORE I sign PDs MEDs Severity Ratiof
Impaired insight 71% 12% *** 14.7 ***
Dull and inattentive 74% 18% *** 10.2 ***
Self-preoccupied 91% 44% *** 8.8***
Slowed speech 74% 24% *** 6.4***
Nihilism 68% 24% *** 6.0***
Non-responsive to interviewer 79% 44% ** 5.2***
Non-reactive to interviewer 97% 97% 5.1**
Facial immobility 94% 80% 3.8**
Endogenous quality 94% 85% 3.7**
Monosyllabic speech 74% 47% 3.6**
Limited mobility 79% 44% ** 3.4**
Slowed movements 74% 41% * 3.4**
Poverty of associations 85% 74% 3.1*
Slumped posture 77% 44% * 3.0*
Signs of anxiety 82% 68% 2.9*
f Odds ratio > 1 indicates over-representation of feature in PDs vs. MEDs.
* P < 0.05; ** P < 0.01; *** P < 0.001.
Adapted from Parker et al. (1991a).

Signs. The total CORE score was 37.1 for the PDs, 25.8 for the EDs and
26.1 for the MEDs. Thus, the PDs scored 40% higher on our initial CORE
measure - one focussing essentially on psychomotor retardation. Table
12.1 data suggest that most of the 15 signs were significantly more likely
to be both present and more severe in the PDs compared to the MEDs.
The PDs distinguished themselves most clearly as evidencing impaired
insight, cognitive processing difficulties (i.e., dull and inattentive; self-
preoccupied) and psychomotor retardation.

Symptoms. We compared the groups on a wide range of symptoms, with


most held to characterise the melancholic or endogenous type of depres-
sion. As anticipated from the rules guiding allocation to the PD group,
delusions and/or hallucinations were the most specific features, with 34
(97%) of the 35 PDs having a delusion (24 mood-congruent) and 7 (20%)
having hallucinations during the episode; such features were absent in
the comparison groups.
A series of univariate analyses established that, compared to the
MEDs, the PDs differed on only a few symptoms. They were
Psychotic Depression 187
(i) more likely to report constipation;
(ii) more likely to acknowledge suicidal plans;
(iii) less likely to show and report diurnal mood variation (i.e., mood
worse in A.M.); and
(iv) more likely to report a sustained mood disturbance on a number of
parameters - which we refined down to four - pessimism about the
future, hopelessness, helplessness and worthlessness.
Trends for the PDs to be more likely to report certain so-called endo-
geneity symptoms (i.e., appetite loss, weight loss, anticipatory anhedo-
nia) were not significant. In addition, the PDs were no more likely to
report insomnia (of any phase), suicidal thoughts, to judge their mood
as less reactive, or to report subjective psychomotor change (either agi-
tation and/or retardation). The last finding is of interest when we have
noted that several psychomotor retardation items which were assessed as
observed signs distinguished the groups. We considered whether several
of the signs (i.e., of psychomotor change) and symptoms (i.e., consti-
pation) might be effects of medication, and therefore examined drug
treatment current at intake assessment. Specific details are provided
in our earlier report (Parker et al. 1991a) of the prevalence and dosage
of differing psychotropic drugs across groups. As the PDs were less
likely to be receiving antidepressant medication and as similar numbers
of PDs and MEDs were receiving antipsychotic medication, we judged
that over-representation of the features of interest in the PDs was likely
to reflect the disorder and not effects of medication; we accept, however,
that such a conclusion can remain only tentative.

Clinical Prediction of a Diagnosis of Psychotic Depression


We pursued clinical definition of psychotic depression in two contrast-
ing ways, variably accepting and not accepting diagnostic decision rules.
First, we used a top-down strategy, presuming for the moment that the
putative diagnostic allocation was a valid gold standard and that all
those, and only those whom we established as having delusions and/or
hallucinations, had psychotic depression. Thus, we respected the cri-
teria-based and clinical diagnostic allocation as the criterion-dependent
variable, with its mandatory requirement of delusions and/or hallucina-
tions, and examined how well any delineated clinical variable set classi-
fied the assigned patients (i.e., 35 PDs vs. 102 EDs).
To that end, logistic regression was used to identify that combination
188 G. Parker, I. Hickie and D. Hadzi-Pavlovic

of dichotomous independent variables that best reproduced cases' scores


on the dichotomous dependent variable, with technical details described
in an earlier publication (Parker et al. 1991a). As the presence of delu-
sions and/or hallucinations was correlated perfectly with our diagnostic
allocation (by definition), it was not appropriate to include psychotic
features in examining the utility of the small number of other global
variables identified in univariate analyses. These variables were coded 1
or 0 as follows:
(i) a total CORE score of 29 or more (identified through mixture anal-
ysis as a cut-off identifying mainly clinically diagnosed psychotic
depressives) as a measure of distinct psychomotor change;
(ii) sustained depressive content;
(iii) the absence of any diurnal variation in mood (staff-observed); and
(iv) constipation.
Clinical assignment of subjects as PD or ED was the outcome vari-
able in the logistic regression analyses. The model fitted the data well
(X2 = 109, df 120, P = 0.75), with only 5/125 patients (all PDs) being
discrepant. Additionally, it provided a significant improvement over the
null model (decrease in deviance = 47.57, df 4, P < 0.0001). Simplifying
the model by deleting the sustained depressive content item resulted in a
non-significant increase in the deviance, with the three remaining items
significant by the approximate t values. We examined the association
between those three variables and the PD class using odds ratios. We
found that patients positive on CORE were 10.8 times more likely to be
a clinically diagnosed PD; those affirming absence of diurnal variation
were 6.2 times more likely; and those positive on constipation were 3.5
times more likely.
Second, we used a bottom-up strategy, no longer assuming our cri-
terion diagnostic allocation to be valid. Here we instead assumed that
two classes of psychotic and non-psychotic melancholia could be iden-
tified statistically. When using categorical variables (here the pres-
ence/absence of a clinical feature), latent class analysis (LCA) is an
appropriate method for this kind of "clustering" (McLachlan 1992) and
was used.
We analysed the four variables listed earlier plus a new variable (HAL-
DEL) which indicated the presence or absence of delusions and/or hal-
lucinations. First, we fitted a one-class latent class model to the 125
available subjects (including 34 PDs), but it showed a poor fit to the
data (X2 = 95.64, df 26, P < 0.0001), arguing against there being only
Psychotic Depression 189

Table 12.2. Latent class analysis of five binary items in combined


psychotic (PD) and endogenous (ED) patients, and correspondence
between latent class solution and clinical diagnosis

Probability of positive response


for member of class
(Standard error)
Variable I II

Hallucination or delusion 0.76 (0.09) 0.00 (0.12)


High CORE score 0.80 (0.07) 0.20 (0.05)
Constipation 0.57 (0.08) 0.29 (0.06)
No diurnal mood variation 0.45 (0.08) 0.15 (0.04)
Sustained depressive content 0.90 (0.05) 0.52 (0.06)

Estimated prevalence 0.36 (0.06) 0.64 (0.07)

Clinical diagnosis n n
Psychotic depressives 34 100% 0 0%
Endogenous depressives 11 12% 80 87%
Total 45 36% 80 64%

Goodness of fit: X2 = 11.2, df 20, P = 0.94.


Test for 2 classes vs. 1: X2 = 84.5, df 6 , P < 0.001.
Adapted from Parker et al. (1991a).

one type. We then fitted a two-class model which both gave a good fit
to the data (X2 = 11.17, df 20, p = 0.94) and was a significant improve-
ment over the fit from the one-class model (see Table 12.2). In the latter
analysis, one class, with an estimated prevalence of 36%, was defined in
one of two ways: either by the presence of HALDEL or by an appropriate
pattern of positive responses to the other four items, but requiring the
definite inclusion of CORE. Thus, the presence of delusions and/or hallu-
cinations was a sufficient, but not necessary, condition for assignment to
this first class (i.e., some subjects judged as not having delusions or hal-
lucinations were allocated to the psychotic class). The other class, with
an estimated prevalence of 64%, was defined by the absence of HALDEL
and a low rate of affirmation of the other four features. When compared
against diagnosis, the solution shows high sensitivity, with 34/34 (100%)
of the PDs being assigned to the first class, but weaker specificity, with
80/91 (88%) of the EDs being assigned to the second class (kappa =
0.80). We then undertook an independent re-examination of the case
files of those 11 patients clinically diagnosed as non-psychotic but al-
190 G. Parker, I. Hickie and D. Hadzi-Pavlovic
located to the psychotic latent class. Such subjects included patients
with "pathological concerns about impending disaster," "hypochondri-
acal concerns about cancer" and a "strange concern about not washing
hair"; chart review established that 3 were definitely delusional in a pre-
vious or subsequent episode. In retrospect, we judged that 9 of these 11
patients may well have had delusional features but failed to express or
acknowledge them with sufficient clarity to meet strict entry criteria to
the PD diagnostic group.

Discussion of the CORE I Study


We need to respect a bias emerging from our sample being derived from
a specialised facility which attracts a significant number of referrals with
chronic and "treatment-resistant" disorders. Such a bias is likely to af-
fect some of our descriptive data (e.g., ratios of psychotic/non-psychotic
melancholies; likelihood of hospital admission; the poor course of sample
members) but, as these biases would be expected to occur equally across
both psychotic and non-psychotic melancholic sub-groups, they should
not have distorted our comparative analyses of clinical features. The
significance of seeking to define psychotic depression clinically and to
distinguish it from non-psychotic melancholia emerges from the specific
finding noted in the introduction - that psychotic depression responds
poorly to antidepressant medication alone, and that there is a distinctly
improved outcome with combination antidepressant/antipsychotic medi-
cation and/or ECT. We therefore pursued a phenomenological definition
of psychotic depression, examining: (i) whether clinical features other
than psychotic ones help define the disorder; (ii) the importance of psy-
chomotor disturbance to the condition; (iii) whether delusions and/or
hallucinations are necessary and sufficient features to a definition of psy-
chotic depression, and (iv) for any evidence that psychotic depression is
a separate type of depression.
We examined a more comprehensive list of clinical features than previ-
ous studies and confirmed the three most frequently suggested findings:

(i) the PDs evidenced greater psychomotor change (as measured by


the CORE score), with the PDs being particularly more likely to
rate as dull and inattentive, self-preoccupied, non-responsive to the
interviewer and showing greater impairment of insight;
(ii) the PDs showed non-significant trends for varying endogeneity symp-
toms to be over-represented; and
Psychotic Depression 191
(iii) while the PDs reported a more severe mood disturbance, it was
distinguished more by being sustained than by severity.

Additionally, we identified two features not suggested in previous re-


views. First, the PDs had a reduced likelihood of experiencing diurnal
variation in mood. We note, however, that there is some clinical support
for this item, with Healy and Williams (1988) referencing a German ob-
servation that "In the severest cases, as patients become progressively
more psychomotor retarded, diurnal variation of mood is often lost."
Also, this feature is included in the ICD-9 definition of reactive depres-
sive psychosis. Secondly, we identified constipation, which has also been
noted in a number of reports. For example, Kraepelin (1907) described
physical symptoms in the "delusional form" where "the appetite is poor,
the tongue coated, and the bowels constipated."
Another objective was, in essence, to examine the necessary and suf-
ficient status of delusions and/or hallucinations in the definition of psy-
chotic depression. Our top-down analyses could not address that issue
- as the presence or absence of such features defined the diagnostic
allocations - but allowed us to test for any superfluousness of the 'psy-
chotic' item to class definition. The four remaining variables established
a moderate predictive potential, but they were insufficient as a set to
act as proxies for psychotic features. Our bottom-up analytic strat-
egy used LCA to isolate allied characteristics best distinguishing one
latent group from the other. We established only moderate agreement
between the LCA solution and the clinical diagnostic allocation of PD
versus ED. As noted, the LCA effectively reallocated 11 clinically diag-
nosed EDs to Class I, the putative psychotic depression class. Here we
face a key issue. Did the dissonant allocation of 8% of the total sample
in that analysis reflect limitations of the multivariate analysis, or did
we miss diagnosing some true psychotic depressives? The latter could
have happened if patients failed to disclose delusions or hallucinations,
if significant psychomotor change prevents their acknowledgment, or if
we rated some true delusions as being non-psychotic over-valued ideas.
Chart review supported the latter explanations, and we therefore have
some confidence in the allocation of cases by the latent class strategy.
The conclusion that delusions and hallucinations were sufficient but
not necessary to allow a diagnosis of psychotic depression is of key clin-
ical importance, for it suggests that restricting a diagnosis of psychotic
depression to depressive disorders only when delusions and/or halluci-
nations have been elicited, as favoured in recent DSM systems, might
192 G. Parker, I. Hickie and D. Hadzi-Pavlovic

be in error - and so creates the risk of inappropriate treatment. We


suggest that psychotic depression can often be "masked" and the di-
agnosis missed if reliance is made on self-reports of the subject. We
argue for a high index of suspicion, in the absence of formally eliciting
delusions/hallucinations, when there is severe psychomotor change - as
suggested in some diagnostic systems such as DSM-III (which included a
depressive stupor criterion).
Finally, we examined the status of psychotic depression. Using our
final five predictor variables (delusions/hallucinations; high CORE score;
sustained depressive content; absence of diurnal variation; constipation)
we showed that a two-class model fitted the data significantly better
than a one-class model. If the latter had provided an adequate account
of the data, then that would have been consistent with a dimensional
view (i.e., psychotic depression is no more than a [more severe] form of
melancholia). Allowing for the caveats discussed in Chapter 4, we argue
that the analyses are consistent with a categorical view of the disorder
and are inconsistent with a simple dimensional view.

The CORE II Study


As detailed in Chapter 6, we subsequently developed the final CORE mea-
sure (with its agitation, retardation and non-interactiveness sub-scales)
in a second sample of 413 depressed patients. This provided an oppor-
tunity to examine the relevance of each scale domain to psychotic de-
pression in some detail, as well as to re-examine a number of issues iden-
tified in the CORE I study (e.g., the possibility that delusions and/or
hallucinations may not always be elicited from those with a psychotic
depressive illness). In this study, reported in greater detail elsewhere
(Parker et al. 1995b), rather than impose our previous rather strict cri-
terion for psychotic depression allocation (i.e., first meeting DSM-III, RDC
and CLINICAL criteria for endogenous depression or melancholia), we as-
signed subjects to a psychotic depression group only on the basis of
meeting DSM-III-R criteria for major depression with psychotic features.
We further differed from our previous study (where we had a compari-
son group of non-psychotic melancholic depressives) by using a number
of comparison groups - DSM-defined melancholia; Newcastle-defined en-
dogenous; and DSM-defined major depressive episode. The last group
was included because most previous studies of the clinical features of
psychotic depression have made comparisons against a major depression
group. Finally, we differed further from our initial study in first requir-
Psychotic Depression 193
ing all sample members to reach DSM-III-R criteria for major depression
and to have been at or close to the nadir of the episode (to exclude those
in partial remission or those whose depression may not have worsened to
a psychotic stage). As in the previous study, we report bottom-up and
top-down analyses. In addition to administering the CORE II measure,
baseline information was collected on a large symptom data set and di-
agnostic allocations (clinical, DSM-III-R, and Newcastle) were made as
detailed in Chapter 6.

Sample. Forty-six (14%) of the 327 subjects met DSM-III-R criteria for
major depression with psychotic features (the PDs). The mean age of
the PDs was 58 years, and 72% were female. The DSM-III-R system
allocated more (N = 148) to a melancholic class than those clinically
diagnosed as having endogenous depression (N = 84) and those scoring
6 or more on the Newcastle Scale (N = 62).

Screening Analyses. Delusions were established in 89% and halluci-


nations in 24% of the PDs, and both had absolute specificity (as expected
by definition) to the PD class. Univariate analyses excluded a number
of symptoms as weighted to either psychotic or melancholic depression,
as they showed similar frequencies across the varying diagnostic groups.
The remaining symptoms tended to be over-represented in the melan-
cholic compared to the non-melancholic classes (and often to have higher
frequencies in the PD class), with a correlation matrix indicating that
some (e.g., appetite loss, weight loss) were strongly associated and could
therefore be amalgamated. In addition to delusions and hallucinations,
we therefore identified nine refined symptoms (appetite and/or weight
loss; mood and/or energy worse in A.M.; anticipatory and/or consumma-
tory anhedonia; terminal insomnia; loss of interest; non-reactive mood;
constipation; deserving of punishment; and feeling sinful or guilty) for
examining for any over-representation to the PD class. The PDs re-
turned significantly higher CORE scores (total = 21.8; retardation =
7.2; agitation = 5.3; non-interactiveness = 9.2) than the DSM-allocated
melancholic (9.2; 3.3; 2.6; 3.4) and Newcastle-allocated endogenous de-
pressives (14.9; 5.5; 3.8; 5.6), indicating significantly greater psychomo-
tor disturbance in the PDs than in the melancholic comparison groups.

Bottom-up Analyses. We first tested our a priori assumption that


the sample had three differing depressive populations (psychotic depres-
sion, melancholia and a residual non-melancholic depression group). We
194 G. Parker, I. Hickie and D. Hadzi-Pavlovic

Table 12.3. Latent class analysis based on all patients (three classes
fitted).

Item probabilities!
Class A Class B Class C
Clinical variable PDt Melt Non-melt
Symptoms
Appetite and/or weight loss 0.94 0.71 0.54
Mood and/or energy worse A.M. 0.49 0.51 0.37
Anticipatory and/or
consummatory anhedonia 0.66 0.57 0.07
Terminal insomnia 0.54 0.47 0.13
Loss of interest 0.70 0.43 0.02
Non-reactive mood 0.39 0.33 0.01
Constipation 0.68 0.43 0.19
Deserving of punishment 0.47 0.20 0.19
Sinful or guilty 0.52 0.22 0.08
Hallucinations 0.13 0.00 0.00
Delusions 0.49 0.00 0.01
Signs
CORE retardation scale 0.92 0.39 0.00
CORE agitation scale 0.75 0.21 0.14
CORE non-interactiveness scale 0.98 0.62 0.13
Prevalence 25% 35% 40%

f For symptoms probability of symptom being present at pathological


level. For signs the probability of being rated 1-3 versus 0.
t Putative classes: PD = psychotic depression; Mel = melancholia;
Non-mel = non-melancholia.
Adapted from Parker et al. (1995b).

therefore undertook a mixture analysis of our provisionally refined set


of clinical features (i.e., delusions and hallucinations; CORE scale scores;
the nine consolidated endogeneity and/or psychoticism items), examin-
ing the total score on all these items. The reduction in the log likelihood
in going from one to two, and from two to three populations was signif-
icant, suggesting that fitting three classes was appropriate.
We then undertook an LCA to estimate the probability of each possi-
ble response for each variable for an individual from the three particular
classes. Table 12.3 reports the three-class LCA solution, with Class A
(containing 82 subjects) being labelled putative psychotic depression,
Class B (N = 115) putative melancholia, and Class C (TV = 130) the
Psychotic Depression 195
putative residual non-melancholic class. It is noteworthy that 25% of
the sample were allocated to Class A, when only 14% of the sample
had been formally diagnosed as having DSM-defined psychotic depres-
sion. In an LCA, item probabilities represent the probability of positive
response to each item by a member of the defined classes. On theoret-
ical grounds, we might anticipate for the putative psychotic depressive
class an item probability approaching 1.0 for delusions but a distinctly
lower item probability for hallucinations, as, while having specificity, hal-
lucinations are uncommon in psychotic depression. Additionally, their
loadings in Classes B and C should be zero if the item has absolute speci-
ficity to psychotic depression. Differential loadings between Class A and
Class B, and Class A and Class C, are required if a feature has some
specificity to psychotic depression as compared to melancholic and non-
melancholic major depression respectively. The item loadings recorded
in Table 12.3 across the three classes showed patterns similar to those
in the prevalence data recorded against DSM- and Newcastle-allocated
sub-groups, supporting our labelling of the putative classes (i.e., psy-
chotic, melancholic and non-melancholic depression). Focussing on the
Class A pattern, only two items (i.e., delusions and hallucinations) had
absolute specificity to this class, but the probability loadings were not
high, suggesting that other features (alone or combined) were defining
this psychotic depression class. CORE scale scores generated the highest
probabilities to Class A, with particularly high loadings from the re-
tardation and non-interactiveness scales. Several other features (feeling
sinful/guilty; sense of deserving punishment; loss of interest; constipa-
tion; appetite and weight loss) were more probable in Class A but also
had reasonable representation in Class B - indicating some discrimina-
tion but not specificity.

An additional perspective on psychomotor change emerges when psy-


chotic depression is defined most broadly (i.e, by the three-class LCA).
In that solution, non-psychotic melancholia was characterised principally
by psychomotor change involving non-interactiveness and retardation. If
we allow for measurement error and/or the overall base rate of anxiety
and/or agitation in all depressives (as indicated by the level of agita-
tion measured in the non-melancholic depressives), then true agitation
appears to be an uncommon feature in non-psychotic melancholies. By
contrast, those with psychotic depression are characterised by a marked
degree of all forms of psychomotor disturbance, but particularly by high
levels of agitation.
196 G. Parker, I. Hickie and D. Hadzi-Pavlovic
Top-down Analyses
We undertook two such analyses: (i) assuming that the DSM-m-R def-
inition of psychotic depression was correct and then (ii) assuming the
LCA Class A definition was correct, seeking to isolate those variables
(other than delusions and hallucinations) which might make a significant
contribution to the two definitions of psychotic depression. We used dis-
criminant function (DF) and logistic regression (LR) to examine the
utility of the remaining 10 variables (9 symptoms and total CORE scores
- the latter consolidated measure being used in the light of significant
Class A loadings on all three composite scales in the earlier LCA).
As each strategy produced similar interpretations, we report odds ra-
tios from the stepwise LRs. The total CORE score was identified as
distinguishing the two variably defined psychotic depression sub-groups
from the several reference groups, whether melancholic or all remaining
subjects. We subsequently used a separate LR analysis to establish a cut-
off CORE score for distinguishing the LCA-allocated PDs from residual
depressives. This score of 22 or more (sensitivity = 48% and specificity
= 96%) was distinctly higher than the cut-off of 8 or more for melan-
cholia established in an earlier analysis. Regarding an odds ratio of 2 or
more as indicating a significant loading to psychotic depression, then a
number of symptoms were also identified as over-represented. Consid-
ering DSM-defined "psychotic depression" first, the following variables
demonstrated specificity, whether the reference group was DSM-defined
melancholia, Newcastle-defined endogenous depression or DSM-defined
major depression:

(i) sense of deserving punishment;


(ii) feeling sinful/guilty;
(iii) appetite/weight loss;
(iv) terminal insomnia; and
(v) constipation.

Considering LCA-defined psychotic depression next, and whether the


comparison groups were LCA-defined melancholia or all non-psychotic
subjects, the same variables (together with loss of interest) were identi-
fied. Thus, whether the DSM or LCA definition of psychotic depression
was considered, a similar list of variables was identified. Using the ini-
tial set of 10 variables (i.e., not including delusions or hallucinations), we
could predict 81% to 93% of the DSM and LCA-allocated PDs and non-
PDs. As the exclusion of delusions and hallucinations as predictors may
Psychotic Depression 197
have distorted the analyses, we repeated the multivariate analyses with
delusions and hallucinations added to the base set of 10 predictors. All
the previously suggested discriminating variables retained their signifi-
cance, and there was little improvement in overall discrimination (e.g.,
the overall correct class allocations were increased from 93% to 94% for
the Class A vs. remainder analysis, and from 90 to 95% for Class A vs.
Class B).

Discussion of the CORE II Study


Respecting the bulk of previous studies, in our top-down analyses we
made comparison against the more generic major depression category
as well as against variably defined melancholic categories. As in the
previous study, we also undertook bottom-up analyses, rejecting (for the
moment) any a priori classification of psychotic depression and derived a
putative psychotic depression class by LCA. In the CORE I study, such
an analysis increased the size of the pre-defined psychotic depression
group by one-third, and included subjects for whom delusions and/or
hallucinations were either lacking or could not be elicited. In this study,
the LCA again increased the pre-defined psychotic depression sample
considerably (i.e., from 46 to 82), adding subjects for whom delusions
and/or hallucinations had not been established, as in the CORE I study.
This is an extremely important finding, suggesting that a significant
percentage of subjects with likely psychotic depression either will not
have delusions or hallucinations or, if they have them, will not allow
such features to be elicited from the clinician - perhaps at times because
of severe psychomotor disturbance.
Our finding that the set of clinical features identified as significantly
over-represented in our variably defined psychotic depressives was al-
most identical across the varying comparison groups is then extremely
important not only in assisting clinical definition of psychotic depression
but in suggesting a set of features that the clinician should consider as
weighting towards a diagnosis of psychotic depression when delusions
and hallucinations cannot be elicited.
Such a phenomenon has a number of important implications worthy of
restatement. If delusions and/or hallucinations were mandatory to the
definition of psychotic depression, there would appear to be little clini-
cal utility in seeking to identify other differentiating clinical features -
as diagnostic predictability could not be further improved. Exceptions
might include situations when such features may be neither acknowl-
198 G. Parker, I. Hickie and D. Hadzi-Pavlovic
edged nor elicited, perhaps when the patient is severely stuporous. The
possibility that delusions and/or hallucinations may not be obligatory
to psychotic depression suggests even more distinct advantages to iden-
tifying other diagnosis-specific or over-represented features, in addition
to challenging the term "psychotic depression." The general consistency
of our analyses (across top-down and bottom-up analyses, and whether
making comparison against all residual subjects or against melancholic
or endogenous depressives alone) was impressive. Equally important,
the relevant variable set remained significant whether delusions and hal-
lucinations were included as two additional predictors or not, indicating
that they made a significant contribution to the prediction above that of
delusions and hallucinations. Each of the features, other than delusions
and hallucinations, may benefit from some consideration.

Psychomotor Disturbance. While our symptom-based data (Parker


et al. 1994) on psychomotor constructs (i.e., items such as slowed down
physically and thinking slowed) did not distinguish the psychotic from
the non-psychotic melancholic classes, behaviourally observed PMD (rated
via the CORE system) was significantly higher in the PDs. We thus
emphasise again the potential dissonance between rating psychomotor
disturbance by subjective and observer reports, and how its suggested
relevance varies according to measurement issues. In the introduction,
we noted that some diagnostic systems allowed assignment of depressed
subjects to a psychotic depression category in the absence of actual psy-
chotic features when depressive stupor was evident. As we established
a cut-off CORE score of 22 or more as predicting psychotic depression in
this study, and recognise (at a clinical level) that a high CORE score is of-
ten associated with profound non-interactiveness (at times extending to
stupor and generating a diagnosis of pseudo-dementia) in depressed pa-
tients, the utility of including severe psychomotor disturbance in some
diagnostic schema as indicative of psychotic depression becomes clear
and is therefore supported. In their review of published studies dis-
tinguishing psychotic from non-psychotic major depression, Glassman
and Roose (1981) emphasised that PMD is "significantly more severe
or more common" in the psychotic depressives, seemingly independent
of mood severity, but with either greater agitation and/or retardation
being variably reported, and that cognitive impairment is somewhat
more common. Paykel (1971), for one, demonstrated both distinctly
greater agitation and retardation in his cluster-analysis-derived "psych-
otic depressive" group. Mean CORE scale scores (and our three-class
Psychotic Depression 199
LCA reported here) indicated that those with psychotic depression dis-
tinguished themselves most clearly from those without non-psychotic
melancholia on the motor (retardation and agitation) sub-scales. While
non-iteractiveness scores were less differentiating, that scale's probabil-
ity loading (0.98) suggests the likely ubiquitous nature of a cognitive
processing problem in those with psychotic depression.

Morbid Cognitions. We consider two of these (sense of deserving pun-


ishment; sinful/guilty) identified in the multivariate analyses, items not
considered in the CORE I study. Both were included in the CORE II
schedule interview in an attempt to tap pathological (but sub-delusional)
expressions of self-reproach and guilt which have been argued from pre-
vious reviews as having some specificity to psychotic depression. Uni-
variate analyses (see Parker et al. 1995b) demonstrated that each was
three times more likely to be reported by a patient who was formally pos-
itive for delusions, but probability loadings in the non-PD latent classes
suggest that a percentage of non-psychotic patients were also likely to
be rated positively, perhaps because both patients and clinicians may
interpret such constructs in quite differing ways (e.g., categorically, di-
mensionally, metaphorically, etc). We suggest that such items approach
or correspond to the concept of pathological guilt but must first be
operationalised on a qualitative basis before attempting to clarify their
relevance to psychotic depression. Such an operational task is unlikely to
be easy, with Aubrey Lewis noting in his scholarly and clinically detailed
doctoral study, that, because of difficulties, it was "better to attempt
no such distinction as between predominant ideas, preconceptions, delu-
sional ideas and delusions" (Lewis 1934).

Terminal Insomnia and Appetite/Weight Loss. These two over-


represented features were also commonly reported by melancholic, and
less commonly by non-melancholic, patients. This graded finding was
most parsimoniously interpreted as reflecting either the greater severity
of psychotic depression and/or the greater severity of any melancholic
base to psychotic depression, as opposed to arguing for any specificity
to psychotic depression.

Constipation. The increased chance of the psychotic depressives re-


porting constipation could be a discriminating clinical feature or a conse-
quence of diagnosis and treatment (with any diagnosed psychotic depres-
sive patient being more likely to have been treated with a psychotropic
200 G. Parker, I. Hickie and D. Hadzi-Pavlovic
medication causing such a side effect). The former interpretation is cen-
tral to some definitions of psychotic depression prior to the introduction
of psychotropic medication (e.g., Kraepelin 1907), and it has been men-
tioned frequently in historical descriptions of "severe depression" over
the centuries (Jackson 1986), where it is also described as "costiveness."
It is of interest to note that a second meaning of "costive" is "slow in ac-
tion or in expressing ideas, opinions," derived from the Latin constipatus.
Thus, the CORE II study suggested a number of variables contributing
to distinctions between psychotic and non-psychotic melancholia. Sev-
eral (i.e., delusions, hallucinations, high CORE score, constipation) were
also identified in the CORE I study, while replication of the remaining
variables identified in the two studies could not have been expected, as
those variables were not included in both data bases.

Conclusions
Both studies identified a key issue: It would appear that despite skilled
clinical assessment, psychotic features may not always be readily elicited
in those with true psychotic depression - whether that is due to either
interviewer limitations or the possibility that psychotic depression can
exist without delusions and/or hallucinations. If the latter is valid, then
any system that includes only psychotic features as the criterion for a di-
agnosis of psychotic depression in depressed patients risks an inaccurate
diagnosis. There may then be some advantages to diagnostic algorithms
providing a default option by including other over-represented features.
Results from these studies clearly indicate that an index of suspicion
would include (i) severe psychomotor disturbance, (ii) over-valued (but
not delusional) ideas concerning guilt and being deserving of punish-
ment, and (iii) constipation preceding any psychotropic medication. The
relevance of several other variables (e.g., a sustained and severe mood
disturbance; lack of diurnal variation; loss of interest) remains to be clar-
ified, but would appear to require operationalised descriptors and per-
haps observational ratings. As a corollary to demonstrating that those
with psychotic depression have pronounced levels of psychomotor distur-
bance (significantly higher than those with non-psychotic melancholia),
we suggest that there is clinical wisdom in suspecting the possibility of
psychotic depression in any depressed patient with severe psychomotor
disturbance, and particularly in those with psychomotor agitation.
Finally, a comment on the status of the disorder. Mixture analy-
sis of scores on the final refined set of items suggested three popula-
Psychotic Depression 201
tions, with the subsequent LCA suggesting that it was reasonable to
label those classes as psychotic depression, non-psychotic melancholia
and non-melancholic depression. This finding was compatible with the
CORE I study in suggesting class distinction between psychotic depres-
sion and melancholia, and in identifying three populations (see Chap-
ter 6) across the two component samples. In our earlier discussion we
conceded that psychotic and non-psychotic depressive disorders could
achieve class distinction merely on the basis of severity (if those lying
at extreme ends of the underlying dimension were selected and con-
trasted), but again sample selection makes that possibility unlikely. If,
instead, psychotic depression truly represents a separate class, then sep-
aration may be further up the taxonomic tree. We now speculate that
psychotic depression may be a sub-class of melancholia - with subjects
having melancholic or endogeneity features (at the same or a somewhat
greater level of severity than those with non-psychotic melancholia), but
additionally having independent sub-typing features (i.e., delusions and
hallucinations in particular).
The differentiation of psychotic from non-psychotic forms of melan-
cholia is likely to be relevant aetiologically. The presence of psychosis
might simply represent the most severe form of melancholia (or of ma-
jor depression), with pathophysiological processes and risk factors being
largely shared. Alternatively, a useful heuristic hypothesis is that psy-
chotic depression involves neurobiological processes different from those
underpinning non-psychotic melancholia. As discussed in Chapter 3,
future research needs to detail clearly the clinical features of psychosis
and psychomotor change, in addition to relevant neurobiological and
psychosocial risk factor data, in order to help clarify these issues. Cur-
rently, we propose that at an aetiological level, psychotic depression
does not occur independently of melancholia. Hence, we hypothesize
that a range of risk factors is common to both melancholia and to psy-
chotic depression (see Chapters 3, 15 and 17). Those with psychotic
depression, however, demonstrate a range of other processes and/or risk
factors which are not shared by those with non-psychotic melancholia.
This view also predicts that true psychotic depression does not occur in
patients with non-melancholic depressive disorders.
Our view contrasts with recent DSM-IV approaches, which have posited
major depressive episode as the stem disorder, in contrast to our postu-
late that the stem disorder is melancholia. Such an explanation concedes
the clear relevance of psychomotor disturbance to both melancholia and
psychotic depression.
13
A Clinical Algorithm for Defining
Melancholia: Comparison with Other
Sub-typing Measures
GORDON PARKER
DUSAN HADZI-PAVLOVIC

Introduction
A number of probability indices or clinical diagnostic criteria sets have
been developed for the diagnosis of melancholia or endogenous depres-
sion, although few have been empirically driven. More commonly, and as
evidenced with recent DSM-IV and ICD-10 criteria sets, the development
process has involved literature reviews and, at the final stages, commit-
tee decisions - relying not only on the utility of descriptors but also on
the imposition of cut-off points for determining caseness. Such a pro-
cess clearly respects clinical wisdom but is sensitive to the composition
of the committee, which may be highly conservative or adventurous,
and at particular risk of preserving the personal clinical views of one
or more powerful committee members. Awareness of such issues pre-
sumably encouraged those developing DSM-IV to weight empirical data.
Thus, in the introduction to DSM-IV (American Psychiatric Association
1994), it is stated that, in arriving at final DSM-IV decisions, the "Work
Groups and Task Force reviewed all the extensive empirical evidence
and correspondence that had been gathered. ... More than any other
nomenclature of mental disorders, DSM-IV is grounded in empirical ev-
idence." The three-stage DSM-IV empirical process included (i) reviews
of the published literature, (ii) reanalyses of available data sets and (iii)
issue-focused field trials.
We also favour an empirically based approach to definition of clini-
cal disorders and here report the development of an algorithm for dis-
tinguishing melancholic from non-melancholic depression. Earlier (see
Chapter 6) we pursued the parsimonious hypothesis that appropriately
defined psychomotor disturbance may be necessary and sufficient to de-
fine melancholia. Analyses suggested that, while psychomotor distur-

202
A Clinical Algorithm for Melancholia 203
bance does define the core of melancholia, several endogeneity symptoms
contribute to the definition. We therefore sought to develop an algorithm
building on those earlier analyses. The utility of that algorithm was then
to be tested by comparative analyses with several current measures of
melancholia and against a number of the clinical and historical givens
about melancholia.

Algorithm Options. We considered several options. One, driven by


the view that melancholic and non-melancholic depression are separate
entities, would require selecting items that best distinguish the two
groups. If based on analyses from Chapter 6, the algorithm might be
restricted to psychomotor disturbance and a small set of endogeneity
symptoms. Such an algorithm would be too parsimonious for many
clinicians, as sub-typing diagnoses are often based on historical rather
than current features, especially if the patient presents when improved
or in remission (when CORE ratings would be invalid). A second, more
dimensional approach (underpinned by a view that melancholia is merely
more severe than non-melancholic depression) might involve developing
a single index with a cut-off score creating two classes that differ by
severity. The risk to that approach is of confounding severity and sub-
typing issues. As the more recent DSM strategy of specifying a number of
prototypic clinical features has been widely accepted, we adopted a sim-
ilar strategy and sought to identify a polythetic set of clinical features.
Finally, respecting the core and mantle model identified for melancholia
in Chapter 6, we imposed a structure allowing a diagnosis of melan-
cholia to be achieved if (i) psychomotor disturbance was present at a
pre-determined level, and/or (ii) a certain number of endogeneity symp-
toms were present.

Comparison Measures. In a DSM-IV commentary, Rush and Weis-


senburger (1994) reviewed nine "current operational definitions" of mel-
ancholia. As our CORE II study subjects had their mood disorder
classified according to DSM-m-R criteria and had been rated on both
the Newcastle Scale (Carney, Roth and Garside 1965) and the 21-item
Hamilton Depression Rating Scale (allowing scores on the derived 8-
item Hamilton Scale for Endogenomorphic Depression [HES] developed
by Kovacs et al. [(1981)] to be generated), we were able to compare our
developed algorithm with three of the systems reviewed by Rush and
Weissenburger.
204 G. Parker and D. Hadzi-Pavlovic

Clinical Validation. As there is no gold standard measure of melan-


cholia, we examined both our derived algorithm and the comparison
measures against a range of clinical givens and against dexamethasone
suppression test (DST) data. A number of reports and reviews (e.g.,
Rush and Weissenburger 1995; Zimmerman et al. 1986b) have suggested
a range of historical and illness course features of melancholia, including
older age, older age at first episode, briefer but more severe episodes,
healthy pre-morbid personality, lack of a precipitant, low placebo re-
sponse rate, a good response to tricyclic medication and to E.C.T., a
poor response to psychotherapy, a higher DST non-suppression rate and
reduced REM latency. We therefore sought data on as many as possible
of these correlates of melancholia.

Methodology. The broad CORE II study design and the sample have
been described in Chapter 6, while the present study has been detailed
in a separate report (Parker et al. 1995c). We note again, however,
that consideration was limited here to those patients who met DSM-III-R
criteria for major depression and who were judged by the clinician rater
as at or near episode nadir. Subjects were alternately categorised as
melancholic or non-melancholic according to (i) DSM-III-R criteria, (ii)
a Newcastle cut-off of 6 or more and (iii) the cut-off of nine or more
established by Kovacs et al. (1981) for the HES. As the latter uses 8
items from the 24-item Hamilton, we took 7 of these from the 21-item
Hamilton and used our clinical rating of hopelessness for the eighth HES
item.
Our data base also allowed us to test diagnostic assignment against a
number of relevant clinical variables noted in the previous section. We
had complete data on most of the clinical and illness correlate variables
but DST data for only a sub-sample - the 85 who were MDU inpa-
tients. As well, we examined for any previous psychotic depressive or
manic/hypomanic episode, anticipating that only melancholic depres-
sives would have had a bipolar illness pattern.
In deriving a list of endogeneity symptoms, we chose the LCA solu-
tion detailed in Chapter 6, as the LCA procedure had considered all
predictors in the set in determining overall class allocation. Assisted
by probability loadings, we selected the following for inclusion in the
algorithm: appetite and/or weight loss, anticipatory and/or consumma-
tory anhedonia, mood and/or energy worse in A.M., terminal insomnia,
loss of interest and non-reactive mood. We did not include the remain-
der for several reasons: constipation because of its unclear status as
A Clinical Algorithm for Melancholia 205

Table 13.1. Agreement between LCA allocation and CORE±SYMPTOM


allocation

Latent Class Allocation^


Algorithm Allocation Mel NMel Kappa Se Sp
CORE> 8 Melf 48 3 0.41 94% 66%
NMelf 67 127
CORE> 8 ± SYMPTOM> 1 Mel 114 112 0.12 50% 95%
NMel 1 18
CORE> 8 ± SYMPTOM> 2 Mel 113 67 0.45 63% 93%
NMel 2 63
CORE> 8 dt SYMPTOM> 3 Mel 108 30 0.70 78% 94%
NMel 7 100
CORE> 8 ± SYMPTOM> 4 Mel 92 6 0.76 94% 84%
NMel 23 124
CORE> 8 ± SYMPTOM> 5 Mel 76 3 0.65 96% 77%
NMel 39 127
CORE> 8 ± SYMPTOM> 6 Mel 66 3 0.56 96% 72%
NMel 49 127

f Mel = melancholia; NMel = non-melancholia; Se = sensitivity; Sp = specificity.


Adapted from Parker et al. (1995c).

either a melancholic/psychotic depression feature or a consequence of


psychotropic medication; and two other items (i.e., deserving of punish-
ment; sinful/guilty) because of low probability loadings in the melan-
cholic class and their greater weighting to the psychotic depression class
(discussed in Chapter 6).

Developing the Algorithm


Our algorithm model had two criteria: (i) a CORE score of 8 or more (Cri-
terion A) and/or (ii) a number (to be determined) of equally weighted
endogeneity symptoms (Criterion B). We proceeded by examining levels
of agreement between affirmation of a progressively increasing number
of the endogeneity symptoms (independent of subjects being above or
below the CORE score criterion of 8 or more) and allocation to the LCA
Class B and Class C cells (which comprised 115 and 130 of those with
putative melancholic and non-melancholic depression respectively).
Table 13.1 reports the cell allocations and levels of agreement with
LCA assignment as we modify the predictive criteria set. Assuming the
LCA allocations to be valid, the ideal cut-off would show close agree-
206 G. Parker and D. Hadzi-Pavlovic
ment between the LCA and algorithm-derived classes, with sensitivity
regarded here as the percentage of LCA-defined melancholies correctly
identified by the criteria set, and specificity being the percentage of LCA-
defined non-melancholies correctly identified, and with overall agreement
quantified by the kappa coefficient. Table 13.1 analyses demonstrate that
entering Criterion A (high CORE score) alone produces an allocation in
modest agreement with the LCA allocation (K, = 0.41), and a base for
considering whether a certain number of symptoms (alone or in addi-
tion to high CORE score) improved agreement, an issue confirmed by
the increases in kappa. The "four or more" symptom solution generated
the highest overall agreement (K = 0.76), had high sensitivity (94%)
and reasonable specificity (84%), and allocated 98 (30%) of the overall
sample to a melancholic class. Thus, we imposed the Criterion B cut-off
as four or more features from our refined endogeneity symptom list.

Comparative Properties of the Algorithm


Next, after deleting the 46 DSM-m-R PDs from the sample, we examined
the comparative capacity of our algorithmic assignments to melancho-
lia versus non-melancholia against independent clinical variables for the
remaining 281 subjects. Our several comparison measures were
(i) DSM-III-R diagnoses of melancholia vs. non-melancholia;
(ii) Newcastle assignment to endogenous vs. neurotic; and
(iii) HES assignment to melancholic vs. non-melancholic
and each of the algorithm's component criteria alone
(iv) a CORE score of 8 or more; and
(v) a SYMPTOM score of 4 or more
Examining algorithmic assignment first (Table 13.2), those allocated
as having melancholia were both currently older and older at first episode,
rated as having a more severe disorder on the Hamilton (but not Zung)
scale, had a briefer current episode of depression, were more likely to
have had a manic episode in the past, and if reporting a previous de-
pressive episode, were more likely to report delusions and/or hallucina-
tions and to have responded to an antidepressant and/or ECT. DST
results (on a total sample of only 37, as the majority of tested subjects
were in the LCA-defined PD group) showed a non-significant trend for
algorithmic-defined melancholies to fail to suppress. When we disaggre-
gated the algorithm into its two components (i.e., positive on Criterion
A Clinical Algorithm for Melancholia 207
A only or Criterion B only), some of the suggested differences (i.e., age,
DST non-suppression) between melancholic and non-melancholic groups
appeared related more to the Criterion A component, suggesting that
behaviourally rated psychomotor disturbance drove greater specificity
than the endogeneity symptoms.
Examination of the several melancholia systems against the number
of significant variables and DST trends indicated that the algorithm
appeared to be performing at least as well as the two comparison systems
(i.e., DSM-m-R, Newcastle) that allow sources of information in addition
to clinical features to contribute to diagnostic allocation, and somewhat
better than the one other clinical feature-based measure (the HES).

Discussion
In seeking to differentiate melancholic and non-melancholic sub-types,
we accepted the view put by Ni Brolchain (1979) and noted in Chap-
ter 6 - that any attempt to classify should restrict consideration "in
the first instance, to mental state items," with aetiological, background,
course of illness and treatment response factors initially excluded (the
respected "nosology necessarily precedes aetiology" view). We selected
some 22 endogeneity symptoms (in addition to assessing psychomotor
disturbance behaviourally) that have been historically considered to de-
fine melancholia. We then excluded (as detailed in Chapter 6) from that
set those features without specificity to the melancholic/psychotic de-
pressive disorders, as several of our previous studies (e.g., Parker and
Hadzi-Pavlovic 1993) have demonstrated how such non-specific features
(which are generally weighted to depression severity) may obscure any
true point of rarity between depressive types and so confound interpre-
tation.
That process, and consolidation of some associated symptoms into
more general descriptors, resulted in a refined set of potential endo-
geneity symptoms as well as CORE measure scores of behaviourally
judged PMD, for entry into an LCA (Chapter 6), with our main objec-
tive being to identify indicators separating members and non-members
of presumed intrinsic or underlying classes, particularly melancholic and
non-melancholic depression, and we used these two LCA-defined classes
as our criterion measure for agreement in developing and testing our
diagnostic algorithm.
In developing the algorithm, we first built in a criterion of clinician-
rated psychomotor disturbance (pre-established as a CORE score > 8).
Table 13.2. Data contrasting melancholic and non-melancholic depressives assigned by six differing systems on a
number of clinical correlates

Algorithm CORE SYMPTOM DSM-III-R Newcastle HES


M NM >8 <8 >4 <4 M NM M NM M NM
Clinical correlate 116 165 88 193 65 216 148 133 62 219 145 136
Age (mean) 51.7 39.0 53.9 39.8 50.2 42.4 50.9 36.8 57.3 40.5 47.7 40.4
Sex (Female) 66% 61% 69% 60% 66% 62% 69% 56% 68% 62% 67% 59%
Baseline depression variables
- Zung (mean) 58.4 57.1 57.6 57.6 59.0 57.1 57.5 57.7 56.4 57.9 58.7 56.5
- Hamilton (mean) 24.0 18.5 24.3 19.2 25.0 19.5 22.9 18.3 24.5 19.7 24.9 16.3
- Enisode duration 59.7 90.0 54.1 88.1 60.7 82.5 65.2 91.1 48.3 85.7 64.2 91.6
History of affective disorder, previous symptoms, and previous treatment response
- Age first mood
disturbance 38.8 25.2 41.2 26.1 40.0 28.7 37.4 23.5 43.5 27.3 34.8 26.6
- Manic attack 15% 4% 16% 5% 17% 6% 14% 3% 23% 5% 12% 5%
- Delusions or
hallucinations 18% 1% 23% 1% 8% 10% 11% 3% 19% 5% 12% 5%
- AD (N = 184) 45% 29% 46% 31% 45% 33% 48% 19% 48% 32% 40% 31%
- ECT (N = 59) 54% 30% 59% 30% 50% 42% 54% 15% 67% 28% 52% 32%

DST non-suppression (N == 59)


- 0800 h 21% 6% 26% 7% 15% 18% 21% 6% 18% 16% 19% 13%
- 1600 h 37% 19% 39% 25% 35% 30% 36% 24% 41% 27% 30% 38%
- 0800 or 1600 h 40% 25% 42% 29% 39% 33% 41% 24% 41% 32% 35% 38%
M = melancholic; NM = non-melancholic.
Means and percentages in bold indicate that the melancholic and non-melancholic depressives are significantly different.
A Clinical Algorithm for Melancholia 209
In Chapter 6 and an earlier paper (Parker et al. 1994) we established
that it agreed with DSM-iii-R-defined melancholia and Newcastle endog-
enous assignment, albeit leaving little residual allocation potential to
historically suggested endogeneity symptoms. The second set of features
(Criterion B) allowed a complementary mantle of endogeneity symptoms
to contribute to the definition of melancholia). By imposing an and/or
criteria set, we nevertheless built to the benefits of both approaches
(i.e., measuring endogeneity symptoms and/or behaviourally rated psy-
chomotor disturbance) and foreclosed on neither.
A key step was to quantify the influence of varying numbers of en-
dogeneity symptoms as providing an alternative criterion for achieving
melancholic status. Thus, we elected not to develop a monothetic set
(where the individual must be positive on all items to receive a diag-
nosis) and adopted a polythetic model (applying a cut-off that even-
tually accepted any four of the six refined symptoms). An additional
advantage to that strategy was that the eventual format approximates
to recent DSM system approaches. In developing the algorithm, we
tested its classification against our earlier derived latent class alloca-
tions to melancholia and non-melancholia. Those allocations might, of
course, have possessed no intrinsic validity, so we were required to test
our algorithmic assignments against other accepted diagnostic systems
in discriminating against correlates of the melancholic condition, to en-
sure that the derived system was not just another untested melancholic
index.

Utility of the Algorithm Those defined by our algorithm as having


melancholia differed from the residual non-melancholic subjects in be-
ing older; being older at first episode of affective illness; having had a
briefer but severer episode of depression; being more likely to have had a
previous manic attack; and, if having had a previous depressive episode,
having been delusional and/or hallucinated, and being more likely to
have responded to an antidepressant drug or ECT.
As such findings are consistent with the published literature on melan-
cholia, the algorithm is supported. Analyses showed clear trends for
higher DST non-suppression rates in algorithm-allocated melancholies,
but each of the three analyses fell just short of formal significance (pre-
sumably due to the small number of subjects). Inspection of the al-
gorithm components failed to indicate that DST differentiation would
be suggested against Criterion B (four or more endogeneity symptoms),
210 G. Parker and D. Hadzi-Pavlovic
while distinct trends were suggested against Criterion A (CORE score
> 8) for the 8 A.M. and 4 P.M. analyses.
Despite being restricted to clinical features, our algorithm performed
as well as DSM-IH-R and Newcastle allocations (of interest when those
systems include features such as treatment response, history, previous
episodes, family history), and suggested superiority to the one clinical
feature-based index (the HES).
As noted earlier, Rush and Weissenburger (1994) reviewed nine cur-
rent operational definitions of melancholia in a DSM-IV commentary. Ex-
amining clinical features, psychomotor retardation was included in all
nine systems; terminal insomnia in eight; diurnal mood variation and
weight loss in seven; psychomotor agitation in six; guilt in five; anhe-
donia, distinct quality, appetite loss and delusions in four; and non-
reactive mood and loss of interest in three. As all but distinct quality
was included in our initial data set, we allowed the opportunity for those
representative features to emerge in developing our algorithm.
Rush and Weissenburger noted a "final option" considered by the
DSM-IV committee - to broaden the DSM-m category with a shorter
list of symptoms. They commented that the " 'core' symptoms of psy-
chomotor retardation, anhedonia, unreactive mood and distinct quality
to mood, would constitute such a list," with that set of symptoms and
signs being supported by (i) correlation with reduced REM latency; (ii)
recurrent across fact or/cluster analytic studies; (iii) relation to a posi-
tive response to ECT or tricyclic antidepressants; and (iv) a tendency
to run in families. They concluded, however, that specific operational
definitions would have to be developed for each item on that list. Our
empirically derived algorithm clearly shows some correspondence with
that suggested refined set.

Conclusion
Our final algorithm appeared as successful as several systems that in-
clude inter-episode and treatment response variables and so may be of
some utility to clinicians. It differs principally from recent DSM criteria
sets in specifying a level of behaviourally judged psychomotor distur-
bance to be accepted, but then shows correspondence with the DSM
systems in defining a number of inclusion symptoms. Of equal impor-
tance to such findings is the process, in that we demonstrate how an
algorithm can be developed in a strictly empirical way.
14
Rating the CORE: A User's Guide
KAY WILHELM

Basic Instructions
This chapter focuses on some of the practicalities of rating individual
CORE items and should be read in conjunction with the measure itself
(see Appendix).

Intended Use of the Instrument. The instrument can be used (i)


to assist with and enrich the normal clinical interview of depressed pa-
tients, and (ii) to sub-type depressed patients into melancholic and non-
melancholic groups for clinical or research purposes. As the instrument
rates observed behavioural nuances, clinical experience with depressed
patients (especially the severely depressed) is required. While training
is necessary to ensure accuracy in research studies, these guidelines may
assist many to a reasonable standard.
It is important to emphasise that the CORE system is not a diagnostic
measure distinguishing (say) melancholia from dementia. It is a sub-
typing system, to be used when a diagnosis of a primary depression has
been made and when division into melancholic and non-melancholic type
is required, or when a probability estimate of melancholia is sought.

Timing of the Ratings During Clinical Interview. It is recom-


mended that CORE ratings do not take place in the first 20 minutes of the
interview. This time period should enable very anxious subjects to settle
and to assist the rater in discriminating between anxiety and agitation.
While no formal tests of diurnal effects have been made, all things being
equal, morning ratings are likely to be better for capturing psychomotor
retardation and agitation, which appear, respectively, to be more severe
and frequent early in the day. In some research studies, it may be useful
to record the time of day and to use the same time for any repeat testing.

211
212 K. Wilhelm

The validity of ratings depends on the interviewer establishing an ap-


propriate environment in which key aspects of attention, interaction and
reactivity can be assessed. In order to judge the principal factor under-
pinning the CORE (i.e., non-interactiveness), the interviewer must strive
to engage the patient in relevant and varied topics. An interview involv-
ing only specific questions aimed at identifying depressive phenomena
or the use of a structured diagnostic instrument makes it difficult to
determine the presence of signs such as shortened verbal responses and
reactivity of mood.
The interviewer should seek to observe "natural" motor movements
(e.g., speed of entry and exit from the interview, ease of movement
from chair to standing). Setting specific tasks such as finger-tapping or
asking the patients to walk down a corridor may also be useful adjuncts,
although such estimates need to respect characteristics (e.g., age and
disability) of the individual that are independent of the depression.
Ratings are behavioural, that is, based on how subjects are observed
to behave at the interview rather than how subjects say they are. This
is not to discourage empathy on the part of the interviewer, but to em-
phasise that, for the purposes of CORE ratings, the interviewer needs to
step back and be an objective observer. The interviewer is free, how-
ever, to clarify with the subject the nature of any feature recorded, and
judgement may be required concerning the appropriateness of responses
in the context of variables such as social background, level of education
and ethnic origin.

The Ratings. Signs are, in effect, first judged to be categorically either


present or absent and, if present, to be graded according to severity, in
terms of either duration (how much of the time they are present) or
degree (how much the specific psychomotor component is evident). A
score of 0 indicates the sign is absent or trivial, while scores of 1 to 3
indicate definite presence with increasing severity.
During development of the measure, we discussed whether a 0 rating
should convey a general concept of normality or the subjects' presumed
usual behaviour. This begs the question of what is "normal" for the
particular subject. We decided that if the behaviour appears normal or
is easily explained by some factor independent of depressive illness (e.g.,
current medical illness) then a 0 rating should be made.
By contrast, ratings of 2 or 3 always imply a definite pathological
behaviour. A rating of 1 also implies pathology but may, at times, be
used when the phenomenon is relatively subtle or unclear. In practice,
Rating the CORE: A User's Guide 213
disagreement at inter-rater sessions was generally of two sorts: first,
whether to rate a 0 or 1, that is, whether the sign was absent or present
to a minor degree; secondly, agreement that the sign was present, but
disagreement whether to rate 2 or 3, that is, how severe the sign was
deemed to be.
A rating of 3 implies severe disturbance. However, we do not have
a standard rule (such as "the subject at his or her presumed worst")
and instead have devised specific descriptors for each item. Some frame
of reference, however, is required; we recommend that the subject be
viewed in comparison with a group at the severe end of the spectrum
rather than "the worst case ever seen." This allows some variation, as,
for example, a rater working in an office psychotherapy practice will
have a different range from a rater based in an acute admitting ward, so
clinicians should consider "worst" in light of the breadth of the known
clinical world, not merely from their individual experience. When in
doubt as to which of two scoring options to select, we recommend rating
conservatively by choosing the less severe option.
There are differences in rating retardation and agitation items, in that
retardation tends to be more consistent across the course of the inter-
view, while agitation is not as sustained and may occur only in episodic
bursts. For retardation items, the instructions therefore recommend rat-
ing the average levels based on an overview of the entire interview, after
allowing for the initial settling-in phase. The agitation items are rated
in terms of their average level of severity even if present for only short
periods of the interview, with duration as a secondary consideration.

Commentary on Individual Items


The conventions for rating the CORE evolved from regular inter-rater
meetings involving MDU clinicians. During the meetings one of the
MDU clinicians interviewed a depressed subject in full view of the group.
The first part of the interview consisted of the structured Hamilton Rat-
ing Scale for Depression, followed by individually relevant open-ended
questions to provide further clarification of depressive phenomena, and
to allow the subject to demonstrate any reactivity of mood, capacity for
social interaction and flexibility of thought processes. The assessment
of individual items was then discussed by all the raters.
The items on the standard CORE rating form are intentionally pre-
sented in random order; however, for this chapter they have been group-
ed according to the three sub-scales discussed in Chapter 6, reflecting
214 K. Wilhelm
truncal non-interactiveness items and arborising agitation and retarda-
tion items. When rating, it is important to note that the constructs of
retardation and agitation, while factorially uncorrelated, are not mutu-
ally exclusive, with depressed subjects positive on agitation items also
often positive on retardation items. Comments pertaining to individual
items should be read in conjunction with the instructions on the rating
form (see Appendix).

The Non-interactiveness Scale Items


These items assess cognitive processing domains, impairment of which
in severely depressed patients may invoke the term "pseudo-dementia."

Non-interactiveness (Item 1). This item captures the extent to


which the subject attends to the interview process and responds ap-
propriately to the social cues provided by the interviewer. The rating
is made in terms of both the quantity as well as the quality of interac-
tion. This item differs from inattentiveness in that the subject does not
necessarily show impaired concentration. Non-interactiveness also needs
to be differentiated from apathy, non-cooperativeness, negativism and
listlessness due to concurrent medical conditions or medication.

Non-reactivity (Item 4). This item rates the amount of emotional


responsiveness over either the course or segments of the interview. Re-
activity may be spontaneous or observed following attempts by the in-
terviewer to cheer the subject up. It is judged in absolute terms and
is not confined to an improvement in mood. Thus, exploration of af-
fectively sensitive topics may lead directly to increased expressions of
distress, which again indicates at least some reactivity. We allow then
for reactivity to go in both directions. While the item principally seeks
to gauge the capacity of a depressed subject to "cheer up," reactivity is
also evidenced by a worsening of mood if relevant within the context of
the interview (for example, a subject who cries when discussing a de-
pressing issue). This is made clear by the instruction for the 3 rating to
be preserved for situations when there is a "severely non-reactive mood
(neither worsening nor improving)."
The notes for the item recommend asking the subject to smile. How-
ever, some non-reactive subjects may respond with a forced or superfi-
cial smile. In the absence of any other evidence of reactivity during the
interview, a slight or transient smile without any apparent depth
Rating the CORE: A User's Guide 215
rates 2. A completely forced smile, without any accompanying warmth
or reactivity, rates 3.
Reactivity is best rated later in the interview when the therapeutic
milieu has been established. It is rated more in terms of presence and
intensity of response rather than its duration. Thus, for someone who
might have appeared non-reactive for an extended period, and who then
gave a rich, deep and natural smile, a 0 rating would be appropriate.
Clearly, the interviewer's inter-personal style and choice of topics should
be set to encourage any intrinsic reactive potential.

Inattentiveness (Item 8). Inattentiveness is based solely on the inter-


viewer's judgement of the subject's inability to attend to the interview
and the interviewer and not on complaints of difficulties in concentra-
tion. The subject fails to appreciate key aspects of the situation and/or
to remain in contact with the flow of the interview, although the subject
may maintain eye contact and even smile at the interviewer. This item
needs to be differentiated from any lack of motivation on the part of the
subject to being interviewed.

Poverty of Associations (Item 12). Here the subject gives responses


that are sparse and without elaboration. This item needs to be differ-
entiated from deliberate defensiveness, obfuscation or vagueness of ex-
pression or concreteness, as a characteristic communication style. An
exploration of a specific topic of keen personal interest to the subject
(e.g., hobby, sport, grandchildren), particularly for those who lack for-
mal education or who are of low intelligence, may aid clarification.

Impaired Spontaneity of Talk (Item 16). This item requires the


interviewer to provide opportunities for spontaneous speech or open dia-
logue (e.g., by asking patients' opinions, views). Spontaneity is assessed
in relation to willingness to initiate conversation and in actual sentence
construction. Education, intelligence and motivation can qualify intrin-
sic levels considerably.

Length of Verbal Responses (Item 7). If the subject is monosyl-


labic or completely mute (a very rare occurrence), a 3 rating is made.
For a 2 rating, each response may consist of only a few words. There
is a need to repeat our general instruction here - that ratings should
occur later in the interview after the subject has had time to warm up.
216 K. Wilhelm

Care must be taken in rating responses to this item during interview


segments using closed questions.

The Retardation Scale Items


We note again that psychomotor retardation varies, both over the period
of an interview and across the day, when it will show diurnal variation;
therefore, morning ratings are favoured. Ratings on these items should
reflect the average level of retardation observed.

Slowed Movement (Item 13). This item measures speed, not total
amount, of movement. While we recommend generally avoiding rating
early in an interview, the speed at which the subject enters the interview
room may give useful early information. A check may be provided by
observing the patient leaving the interview room.

Facial Immobility (Item 2). Raters should focus on the degree of


expressiveness (particularly around the eyes) and fluidity of facial move-
ments. The presence of a full range of facial movement rates 0, even
if present for only some of the interview. Facial movements should be
observed closely as changes of topic are introduced.

Body Immobility (Item 10). This item measures amount and de-
gree of body movement rather than speed. As some relatively immobile
patients may still gesticulate to a degree, ignore gesticulation in your
ratings. Some anxious subjects tend to freeze, particularly early in the
interview. Our general instructions about delaying ratings and relaxing
the patient assist here.

Postural Slumping (Item 3). Rating of this item can be influenced


by the relative position of interviewer and subject, as slumping is easier
to see when the interviewer or observer is viewing the subject from the
side. It is important to note whether the subject is using the armrests
on a chair for support, and also to observe the subject while he or she
is standing and, in particular, walking.

Delay in Motor Activity (Item 15). This item measures the dura-
tion of delay in initiating voluntary movement. It is often best rated
when the subject gets up from the chair to leave the room. Otherwise
the subject should be asked to perform a specific motor task.
Rating the CORE: A User's Guide 217
Delay in Responding Verbally (Item 6). This item measures du-
ration of delay in initiating speech (also known as pause time). Verbal
content may be appropriate but delayed. As the extent of delay can
vary, tests of concentration can be used to standardise responses (e.g.,
requesting the subject to give the months of the year in reverse order).
It is important to ensure that the subject has adequate hearing and is
attending to the content of the interview.

Slowing of Speech Rate (Item 17). This item measures the rate of
execution of speech. The considerable variation in the normal population
needs to be respected in rating this item.

The Agitation Scale Items


Agitation has some characteristic features (detailed in the general guide-
lines to the CORE measure) and is often difficult to distinguish from
anxiety, particularly as agitated patients are generally anxious, while
anxious patients may or may not be agitated. Agitation implies an in-
creased amount of movement that tends not to be goal-directed, and is
associated with physical restlessness and increased emotional arousal.

Facial Apprehension (Item 5). The amount of apprehension is rated


once the subject has been given time to become comfortable with the
interview situation. This sign is best observed in the upper half of the
face, particularly around the eyes, with any worried, frightened appre-
hensive mien typically little affected by time spent with the interviewer.
This item should be differentiated from (i) a worried frown, which is
more apparent in the corrugator muscles of the forehead and which will
tend to diminish with reassurance or change of topic; and (ii) perplexity,
where the subject appears puzzled by particular questions or tasks.

Facial Agitation (Item 9). This sign should be rated more in terms of
its intensity (when present) rather than its persistence. It often appears
in epochs seemingly unrelated to the interview situation or, at times,
obviously connected to distressing themes emerging or being explored
at interview. Facial agitation is again usually more evident or marked
in the upper half of the face. Ratings of agitation should be distinguished
from tardive dyskinesia or other involuntary movements (such as tics),
or movements associated with poorly fitting dentures.
218 K. Wilhelm
Motor Agitation (Item 11). This sign should also be rated more
in terms of its intensity (when present) rather than its persistence or
periodicity. The instructions for this item concentrate on movements
characteristic of mental perturbation. Ratings should focus on motor
components and should not be at all influenced by subjective distress.
Gesticulations and involuntary movements (e.g., tics, myoclonic jerks,
and extra-pyramidal dysfunction) do not generate positive ratings.

Stereotyped Movements (Item 18). All subjects with stereotyped


movements will rate on motor agitation items, but not all agitated sub-
jects will rate on stereotyped movements. If motor agitation (Item 11)
has returned a 0 rating, then a 0 rating must be returned here. Motor
stereotypies may be constant, may occur in epochs seemingly unrelated
to the content of the interview or may be increased when the subject is
addressing a distressing theme or issue.

Verbal Stereotypy (Item 14). This item is the verbal equivalent of


stereotyped movements and evidenced in repetitive speech within a re-
stricted range of expression. The important feature here is the repetition
of the same phrases rather than the actual content, although despair and
importuning are common themes. This item should be distinguished
from perseveration of speech (where a response that was appropriate to
the first stimulus is repeated when it is no longer relevant to subsequent
stimuli), and a restricted verbal range due to developmental delay or
limited education.

Do Standardised Motor Tasks Assist CORE Ratings?


Over time, we have considered whether it is helpful to use set motor and
verbal tasks to allow for comparison of motor speed and speech rate. At
times, we have asked subjects to finger-tap, hand-clap or pretend to use
a hammer, but we believe that intrinsic motor skills, age and personality
characteristics (e.g., being careful to undertake the task exactly right or
not make a mistake) restrict their potential utility. As yet, we have failed
to identify any standardised motor task that assists clinician-generated
CORE ratings.
Rating the CORE: A User's Guide 219
Learning to Use the CORE
A training videotape with a step-by-step discussion of each item and
illustrations of patients rating positive and negative was made in 1991.
It is used in our training sessions but is not distributed to other centres
or researchers, to respect the permission releases contracted with the
videotaped patients. If researchers or clinicians wish to be trained to an
acceptable standard, our training procedure generally involves (i) view-
ing and discussion of the videotape in formal training sessions, and (ii)
rating videotaped patients or patients at interview, with ratings checked
against the training psychiatrists' ratings until adequate reliability is
demonstrated.
While an MDU rating session is recommended as the most effective
method of instruction, it is hoped that these notes and the general in-
structions on the measure will suffice where formal training is not pos-
sible.
Part Three
The Neurobiology of Melancholia
15
Melancholia as a Neurological Disorder
MARIE-PAULE AUSTIN
PHILIP MITCHELL

Introduction
The last few years have witnessed a major paradigm shift in the un-
derstanding of basal ganglia function. In line with this has been an
increasing appreciation of both the centrality of psychomotor deficits in
melancholia and the striking clinical parallels between melancholia and
certain basal ganglia disorders. In view of these developments, a more
considered appraisal of the role of the basal ganglia and their frontal
connections in the pathogenesis of melancholia has become possible
(Krishnan 1993b; Austin and Mitchell 1995).
There is now a large body of work focussing on the presence of depres-
sion in a number of neurological disorders, particularly those affecting
striatal and frontal structures (see Starkstein and Robinson 1993 for
review). Because of the historical dichotomy between psychiatric and
neurological disorders, few have explored the possibility that melancho-
lia may be a neurological disorder in its own right, and that the mech-
anisms leading to the motor, cognitive and mood changes in certain
neurological disorders may also be involved in creating a similar, albeit
characteristically individual, combination of such features in melancho-
lia.
As noted in previous chapters, amongst a number of statistical at-
tempts to separate melancholic and non-melancholic depression, some
factor analytic studies have demonstrated that psychomotor retardation
may be one of the best discriminators of these two syndromes (Nel-
son and Charney 1981), a finding extended by our development of the
CORE measure and study of its properties. This chapter respects the
increased interest in the objective or behavioural aspects - both mo-
tor and cognitive - of melancholia. In particular, motor disturbance is

223
224 M.-P. Austin and P. Mitchell
gaining increasing attention as a potential core behavioural manifesta-
tion of melancholia, as reviewed in earlier chapters. Cognitive deficits
are also a prominent feature of melancholia, especially in elderly and/or
psychotically depressed patients, where the impairment is sometimes of
such magnitude that these patients have been described as suffering from
a "depressive pseudodementia" (Kiloh 1961) or "depressive dementia"
(Stoudemire et al. 1989).
Concurrent with this focus upon the behavioural aspects of depres-
sion has been the development of the concept of "subcortical dementia."
This term was introduced by Albert, Feldman and Willis (1974) to de-
scribe the clinical presentation of disorders with predominant subcortical
pathology (such as progressive supra-nuclear palsy, Parkinson's disease
[PD], Huntington's chorea [HC] and Wilson's disease), which are char-
acterised by slowed mentation and movement, apathy, depression, and
reduced ability to manipulate acquired knowledge in the absence of so-
called cortical deficits such as apraxia, agnosia and dysphasia. In view
of the significant "frontal" executive deficits also noted in these disor-
ders, the term was subsequently revised by Albert (1978) to "frontal-
subcortical dementia."
Although the clinical validity of these terms has not always been
self-evident, they have led to an important conceptual shift: that is,
from defining functional psychiatric disorders, including melancholia, in
purely phenomenological terms, to viewing them in terms of regional
cerebral dysfunction and disruption of neuronal networks.
In order to pursue the thesis that melancholia may be a neurologi-
cal disorder in its own right, we shall review the psychomotor deficits
of melancholia and highlight the similarities between these deficits and
those seen in two of the most common, and best studied, of the dis-
orders affecting the basal ganglia and their connections - Parkinson's
disease and Huntington's chorea. Finally, using the clinical parallels
between these three disorders as our springboard, we shall develop a
neuroanatomical and neurochemical formulation of the pathogenesis of
melancholia incorporating the concepts of Alexander et al. (Alexander,
DeLong and Strick 1986; Alexander, Crutcher and DeLong 1990).

Coexistence of Depression with Parkinson's Disease and


Huntington's Chorea
Depression occurs with an increased frequency in patients with disor-
ders that involve basal ganglia pathology. Thus in PD, studies using
Melancholia as a Neurological Disorder 225
the DSM-III-R definition of depression report a mean prevalence rate of
depression of 43% with a range of 20-90% (Cummings 1992) and an
annual incidence rate of 1.86% (Dooneief et al. 1992). While some in-
vestigators (e.g., Gotham, Brown and Marsden 1988) have suggested
that depression in PD may be a psychological reaction to disability,
the greater incidence of depression in PD patients when compared to
controls with equal or greater physical impairment (Mayeux 1990), the
emergence of depression before the onset of motor dysfunction (Mayeux
et al. 1981; Santamaria, Tolosa and Valles 1986) and the lack of associa-
tion between the severity of PD and the presence of depression all argue
against a merely reactive aetiology for depression in this disorder.
In Huntington's chorea, a rate of major depression of 28% is reported
(Folstein 1989). This community study of HC patients highlights that
depression is more common in certain families, in white Caucasians and
in those with older age at onset and may in many cases precede the onset
of motor symptoms by an average of five years. All these factors would
mitigate against depression occurring in these HC subjects simply as a
reaction to their physical disabilities. Phenomenologically, the depres-
sion seen in both PD (Starkstein and Mayberg 1993) and HC (Peyser
and Folstein 1993) patients is strikingly similar to that of idiopathic de-
pression. In addition, the natural course of the depressive illness, suicide
rates and response to antidepressant treatment is similar to that seen in
idiopathic depression.

Psychomotor Dysfunction in Parkinson's Disease,


Huntington's Chorea and Melancholia: Some Parallels
Because of the long-standing historical dichotomy between psychiatric
and neurological disorders, the similarities that exist between the motor
manifestations of melancholia and particular aspects of PD and HC have
rarely been highlighted.

Motor Dysfunction
Whilst PD is manifested by rigidity, tremor, postural instability and
bradykinesia, it is this latter feature which is considered to be the most
characteristic feature of this disorder (Marsden 1989). Similarly, whilst
the choreiform movements of HC are most prominent in the early phase
of this disease, bradykinesia is the most consistent feature of this disor-
der over time (Folstein 1989). The bradykinesia seen in both of these
226 M.-P. Austin and P. Mitchell
neurological disorders bears a striking resemblance to the motor slowing
or retardation so characteristic of melancholia. Despite the prominence
of motor slowing in melancholia, the tools for assessing its extent have
been lacking. In an attempt to do this, Rogers and colleagues (1987)
employed a scale used in the assessment of motor impairment in Parkin-
son's disease - the Webster scale (Webster 1968) - and found that the
scores obtained in melancholic subjects were significantly different from
control scores, although not of the magnitude found in PD.
Parkinson's disease and HC patients also frequently exhibit a slow-
ness of thought, or bradyphrenia, that is indistinguishable from that
seen in melancholic patients (Rogers et al. 1986) and which is a key fea-
ture of the "frontal-subcortical" dementias. Indeed, Hart and Kwentus
(1987) claim that "perhaps the most characteristic clinical feature shared
by depression and certain subcortical neurological disorders is cognitive
and/or psychomotor slowing." Psychomotor slowing, both motor and
cognitive, has been investigated in subjects with PD (Wilson et al. 1980;
Evarts, Teravainen and Calne 1981), HC (Huber and Paulson 1987) and
melancholia (Cornell, Suarez and Berent 1984; Parker et al. 1994), by
means of reaction time (RT) studies. Each of these studies undertaken
on single patient groups reports slowed performance on both components
of the RT task. Studies comparing melancholies to PD (Rogers et al.
1986; Hart and Kwentus 1987) and HC (Wolfe et al. 1987) subjects are
less clear-cut. Thus, while Rogers et al. (1986) reported equivalent pro-
longation of both motor and cognitive components of the Digit Symbol
Substitution Task in melancholies and in PD subjects, Hart and Kwen-
tus (1987) demonstrated that although slowing of motor response is seen
in melancholies, these, unlike PD subjects, appear to have a normal rate
of information processing.
There is a similar lack of consensus in studies looking at more complex
tasks of simultaneous motor function in PD and melancholic subjects.
Thus, Sachdev and Aniss (1994) reported no difference in the rate of
slowing of performance on simultaneous tasks in motor-retarded melan-
cholies and PD patients, whilst Fleminger (1992), claimed to be able
to distinguish between motor-retarded melancholies and PD subjects on
the basis of differential performance on these simultaneous tasks.

Cognitive Dysfunction
Depression. It is now commonly accepted that depression is associated
with a number of cognitive deficits (Miller 1975), especially in the areas
Melancholia as a Neurological Disorder 227
of memory, attention and speed of processing, although it is not always
clear whether such deficits are specific to certain depressive sub-types
or relate to severity. Early controlled studies (Cronholm and Ottosson
1961; Stromgren 1977; Sternberg and Jarvik 1976) found that depressed
subjects used ineffective strategies for initial acquisition and subsequent
retrieval of verbal material from short-term memory stores. When struc-
ture or organisation was provided, the deficit was reversed (Weingartner
et al. 1984). Subsequent studies by the same group (Cohen et al. 1982;
Roy-Byrne et al. 1986), looking at the effect of task difficulty on per-
formance, led to the hypothesis that depress!ves performed poorly on
"effortful" as opposed to "automatic" verbal learning tasks requiring
simpler retrieval strategies. This hypothesis was partially validated by
the finding that recall was more impaired than was recognition of ver-
bal material (Roy-Byrne et al. 1986; Wolfe et al. 1987). Brand, Jolles
and Gispen-de Wied (1992) and Austin et al. (1992b), however, found
impairment on both memory recall and recognition, thus putting into
question the validity of this explanatory model.

Impairment on frontal tasks is also seen, at least in the more severely


depressed subjects (Cassens, Wolfe and Zola 1990). This is manifested
by impairment on the Stroop task (Raskin, Friedman and Di Mascio
1982); the Shift test (Friedman 1964); verbal fluency and the Wiscon-
sin Card Sorting Test (WCST) (Pendleton Jones, Henderson and Welch
1987); and a task looking at self-generated strategy in problem solv-
ing (Silberman, Weingartner and Post 1983). Austin et al. (1992b),
using a broad neuropsychological battery in a sample of patients with
endogenous and non-endogenous depression, found that while all sub-
jects demonstrated memory impairment, only those with endogenous
depression were impaired on tasks involving shifting of cognitive set and
visuomotor ability (Digit Symbol Substitution [from the WAIS-R] and
Trails B [from the Army Individual Test Battery]), indicating that a dif-
ferentiation between sub-types of depressed patients could be made on
the basis of patterns of cognitive deficits. Although psychomotor slow-
ing, present in at least some endogenous patients, may have accounted
for these deficits, relatively intact performance on Trails A (a measure of
visuomotor speed) makes it less likely that this alone accounted for the
difference. In depressed subjects, Channon, Baker and Robertson (1993)
have more recently found impairment in a task of working memory. The
relevance of the working memory paradigm (Baddeley and Hitch 1974)
is that it is derived from animal work, where it is seen as a marker of dor-
228 M.-P. Austin and P. Mitchell
solateral prefrontal circuit function (Alexander et al. 1986), as assessed
by the delayed response task (Goldman-Rakic 1991).

Parkinson's Disease. Impairment is found in a range of cognitive


functions in patients with PD - in particular on tasks of executive func-
tion, verbal fluency, visuospatial skills and memory (see Boyd et al. 1991
for a review). Consistent impairment on tasks of executive function such
as the WCST, and verbal fluency (Lees and Smith 1983; Flowers and
Robinson 1985; Butters et al. 1978; Matison et al. 1982; Taylor, Saint-
Cyr and Lang 1986) and visuospatial skills (Torack and Morris 1988),
are seen in early and untreated PD (Lees and Smith 1983; Butters et al.
1978; Cooper et al. 1991), indicating that these are not secondary to
the effects of medication nor to the more widespread intellectual impair-
ment found in the later stages of PD. It is noteworthy that PD patients
perform significantly worse on the WCST than patients with dementia
of the Alzheimer type when the two groups are matched for severity
of intellectual decline (Pillon et al. 1986), highlighting the preferential
decline in tasks of frontal lobe function.
Impairment of recent memory (Pirozzolo et al. 1982; Huber, Shut-
tleworth and Paulson 1986) is present, with performance on tasks of
verbal recall (Weingartner et al. 1984) being similarly poor in compari-
son to depressed subjects. While impairment of recent memory appears
to be associated with both retrieval and encoding difficulties in these pa-
tients (Caine, Ebert and Weingartner 1977; Josiassen, Curry and Man-
call 1983; Huber and Paulson 1987), as in depression, it is the retrieval
of newly acquired information which is most impaired.

Huntington's Chorea. The range of cognitive deficits seen in HC is


similar to that seen in PD (see Brand and Butters 1986, for a review).
Deficits in cognitive speed and ability to change cognitive sets are par-
ticularly prominent early in the course (Folstein 1989) and often precede
the motor phenomena (Hayward, Zubrick and Hall 1985). As in depres-
sion, recognition for verbal material is relatively spared (Butters et al.
1985; Moss et al. 1986), while verbal recall and verbal fluency are sig-
nificantly impaired (Wolfe et al. 1987). Performance on visuospatial
tasks and those of executive function is also abnormal, even early in the
course when this cannot be attributed to general intellectual decline or
poor motor coordination (Josiassen et al. 1983).
Melancholia as a Neurological Disorder 229
Impairment of the Putative Frontal-subcortical Neural
Networks: Can it Explain the Behavioural Deficits Seen in
Parkinson's Disease, Huntington's Chorea, and Melancholia?
Traditionally, the basal ganglia were considered to be involved only in
the modulation of movement. More recently, however, Alexander et al.
(1986) have proposed the existence of a number of parallel functional
neural networks originating in various regions of the prefrontal cortex
and passing through the basal ganglia which may have an impact not
only on motor function but also on cognitive processes and mood. For
example, motor disturbance could result from dysfunction of Alexander
et al. 's proposed motor loop linking the supplementary motor area to
the putamen. Mood and motivational disturbances seen in these disor-
ders could be related to dysfunction in the limbic loop linking medial
prefrontal structures with the ventral striatum, and cognitive deficits
to disturbances in the prefrontal loop linking lateral prefrontal struc-
tures to the caudate nucleus. (Krishnan [1993b] has also discussed the
potential role of Alexander's circuits in depression in the elderly.) An
important aspect of these parallel functional loops is that similar be-
havioural changes may occur as a result of pathology at various points
in the circuit. Thus, behavioural changes associated with subcortical
lesions will resemble those occurring with frontal lobe dysfunction.
The major evidence for the hypothesised circuits of Alexander et al.
(1986) comes from primate studies suggesting hitherto unrecognised
functional diversity in the basal ganglia. Specifically, these investigations
have examined the behavioural correlates of both single cell recordings
from basal ganglia neurons (Goldman-Rakic 1991; Alexander and De-
Long 1985) and experimental lesions, in both the basal ganglia and the
frontal lobes (Divac, Fonnum and Storm-Mathisen 1977). Lesions in ei-
ther part of the brain lead to a similar inability in the primate to solve
delayed response tasks. These deficits in behaviour seen in primates
are equivalent to those seen in humans with subcortical dementia. Ad-
ditional autoradiographic (Kunzle 1975; Kalil 1978) and histochemical
studies (Graybiel and Ragsdale 1978) of the connections between basal
ganglia and cortical regions also strongly support the existence of frontal-
subcortical pathways. Specific evidence for the existence of a prefrontal
(cognitive) loop is found in studies of primates (Alexander, DeLong and
Strick 1986) and HC subjects (Stuss and Benson 1984), demonstrating
that the observed cognitive deficits (executive function, verbal recall,
230 M.-P. Austin and P. Mitchell
set-shifting and information processing) are served by those parts of the
frontal lobe that provide major cortical input to the caudate.
Using PD as an example, one can see how progressive disruption in
a number of these functional loops could explain the sequence of clin-
ical events. In the initial phase of the disease, PD is associated with
reduced striatal dopaminergic input to the putamen and thus presum-
ably with dysfunction of the motor loop (Alexander et al. 1986). As
the degenerative process progresses, and prefrontal cortex and caudate
ganglia become involved, cognitive and mood changes become manifest
corresponding to involvement of prefrontal and limbic loops.

Identifying Abnormal Regions and Networks in Parkinson's


Disease, Huntington's Chorea and Depression: In Search
of the Putative Neuroanatomical Basis of Melancholia
From the above overview, it is clear that depression frequently occurs in
PD and HC patients, that the bradykinesia of melancholia is indistin-
guishable from that seen in PD and HC, and that a distinct subcortical
pattern of cognitive deficits is common to all disorders. These obser-
vations lead us to pose the following question: What evidence is there
from neurobiological studies for the existence of shared pathogenetic
mechanisms leading to a common phenotypic presentation?

Neurocognitive Studies
In terms of behavioural neuroanatomy, both the impaired executive and
set-switching tasks, as well as the memory deficits seen in all three disor-
ders, may be understood in the broader context of dysfunction in infor-
mation processing, which relies on an intact prefrontal cortex (PFC).
Shallice (1988), a major researcher in this area, postulates that the
PFC acts as a "supervisory attentional system" (SAS), which is con-
sidered critical for the performance of novel or "willed acts" requiring
self-generated problem-solving strategies (as opposed to routine or auto-
matic tasks which do not). Deficits in this SAS, affecting the effective-
ness of information processing and willed acts, have now been postulated
to occur in parkinsonian patients (Morris et al. 1988; Karayanidis 1989)
and in HC (Caine, Ebert and Weingartner 1977) patients, thus providing
a theoretical basis for the frontal neuropsychological deficits seen in this
group. Given the presence of performance deficits in tasks heavily re-
liant on self-generated strategies in all three groups, this same paradigm
Melancholia as a Neurological Disorder 231
may be applied to our understanding of the neuropsychological deficits
found in melancholic depression.
Whilst there is little correlation between motor disability and cogni-
tive impairment in PD (Rafal et al. 1984; Mortimer et al. 1982; Cooper
et al. 1991), cognitive deficits tend to be associated with concurrent de-
pression. For example, Starkstein et al. (1989), in a matched control
study examining cognitive impairment in PD patients with and without
depression, found that the depressed PD patients performed significantly
worse than the non-depressed PD subjects on all aspects of neuropsy-
chological function, but especially frontal tasks. The pattern of deficits
seen in the depressed PD group was strikingly similar to that reported
in Austin et al.'s (1992b) endogenously depressed patients. Starkstein
et al. (1989) went on to hypothesise that both cognitive impairment and
depressive disorder in PD might be associated with pathological changes
in dopaminergic neurons located in the ventral tegmental area, an inte-
gral component of the mesolimbic network. This suggestion is consistent
with the post-mortem findings of Torack and Morris (1988), who noted
marked loss of these neurons in patients who had previously been clin-
ically assessed as presenting with parkinsonism, cognitive deficit and
depression. Conversely, the apparent lack of association between motor
and cognitive manifestations of PD may be understood on the basis of
differential involvement of motor and prefrontal loops (Alexander et al.
1986).

Neurochemical Studies
Neurochemical pathways also undoubtedly contribute to the pathogen-
esis of this clinical triad in all three disorders via their modulatory ef-
fect on the functional networks. In animal studies, manipulation of
dopamine levels in the cortex, striatum and ventral tegmental area have
been shown to affect performance on delayed response tasks, the per-
formance of which in primates is reliant on intact frontal input to the
caudate ganglia (Brozoski et al. 1979; Simon, Scatton and Le Moal 1980).
The well-documented presence of dopaminergic deficits in PD provides
us with a model to look at the potential contribution of dopaminergic
dysfunction to each of the components of our clinical triad. While it
is well recognised that impairment in striatal dopaminergic function is
primarily responsible for the motor deficits seen in PD, it is less clear
whether this neurotransmitter system is also involved in the production
of the associated cognitive deficits. The review paper by De Keyser, Her-
232 M.-P. Austin and P. Mitchell
regodts and Ebinger (1990) highlights the different dopaminergic path-
ways which may be affected in PD and hypothesises that the
mesolimbic pathway connecting dorsolateral prefrontal structures with
ventro-medial caudate areas (and thus corresponding at least in part to
Alexander et al. 's prefrontal loop) may be the one implicated in cog-
nitive deficits in PD. The use of dopamine agonists in the treatment
of PD has allowed study of their effect on cognitive function in that
disorder. Gotham et al. (1988) examined the impact of fluctuations of
brain dopamine levels on frontal lobe task performance in PD patients,
by testing them in both the "on" and "off" phases of L-Dopa treatment.
Variable test results were obtained. Verbal fluency was found to be im-
paired in the "off" state and improved in the "on" state, suggesting that
performance on this task was dependent upon brain dopamine levels. In
contrast, performance on the WCST was impaired in both states, sug-
gesting that either the impairment was independent of dopamine levels
or the test was sensitive to disruption in a number of transmitter systems
and/or brain regions.
The biochemical studies of retarded melancholia and PD (van Praag
et al. 1975; Willner et al. 1991; and see Kapur and Mann 1992 for a
review) suggest a reduction of dopamine activity in striatal pathways,
which in turn is likely to have an impact upon Alexander et al. 's pro-
posed motor loop through the connections between the substantia nigra
and putamen (an integral part of the motor loop).
In normal subjects, dopamine has been shown to facilitate selectively
effortful memory processing in contrast to a more automatic memory
task which was unaffected (Newman et al. 1984). In depressed patients,
the use of dopamine agonists has similarly been shown to improve perfor-
mance on effortful memory tasks. When depressed patients were given a
single dose of dextroamphetamine, improvement was seen on free recall
of verbal material (effortful task), but not on verbal recognition (auto-
matic task) (Reus et al. 1979). Similarly, Murphy, Henry and Weingart-
ner (1972) and Henry, Weingartner and Murphy (1973) found marked
improvement in free recall and a serial learning task in depressed patients
treated with 800 mg L-Dopa for several days. This latter finding was in-
dependent of the mood state, which remained unchanged. Both tricyclic
antidepressants and particularly ECT have been reported to be effec-
tive in the treatment of motor symptoms in PD patients (Strang 1965;
Laitinen 1969; Anderson et al. 1980). This may relate to the dopamine-
enhancing activity of both these treatments (Halaris, Belendiuk and
Freedman 1975).
Melancholia as a Neurological Disorder 233
In conclusion, there is now evidence (Krishnan et al. 1992; Strange
1993) to suggest that dopamine is reduced in both melancholia and PD,
and that this might impact on the behavioural and affective correlates of
these disorders, through its modulatory input into the functional loops
postulated by Alexander et al. (1986).

Imaging Studies
Findings from a small number of neuroimaging studies in depressed pa-
tients suggest that the anatomical regions thought to be involved in
Alexander et al. 's (1986) putative functional networks might be im-
plicated in the pathophysiological mechanisms underpinning the psy-
chomotor deficits of melancholia.

Depression
The introduction of functional imaging techniques, such as positron em-
ission tomography (PET) and single photon emission computerised to-
mography (SPECT), to quantify changes in regional cerebral blood flow
(rCBF) and metabolism holds the potential to elucidate the neuroan-
atomical substrates of melancholia. Studies specifically investigating
patients with more severe forms of depression - endogenous, melan-
cholic and psychotic - have been infrequent and at times conflicting in
their findings. There is a general consensus, however, that in depressed
patients, rCBF and glucose metabolism are reduced predominantly in
subcortical (Buchsbaum et al. 1986; Baxter et al. 1989) - and caudate
in particular - as well as prefrontal areas (Austin et al. 1992a; Bench
et al. 1992; Delvenne et al. 1990). Of particular note is the finding that
melancholic subjects with psychomotor retardation show preferential re-
duction in dorsolateral prefrontal rCBF (Bench et al. 1993b).
Magnetic resonance imaging (MRI), with its ability to provide sharp
resolution of both cortical and subcortical structures, provides a useful
adjunct to the investigation of the anatomical correlates of melancho-
lia. Recent studies have highlighted the existence of a sub-group of
older melancholic subjects with associated deep white matter, striatal
and periventricular hyperintensities (Coffey et al. 1990; Coffey et al.
1993). Furthermore, Brown et al. (1992), in a controlled study of pa-
tients with a variety of psychiatric disorders, found significantly more
frequent and severe hyperintensities only in depressed subjects over the
age of 45. Although the pathophysiological significance of these lesions
234 M.-P. Austin and P. Mitchell
is not clear, it is possible that they may reflect microvascular lesions dis-
rupting functional networks such as those described by Alexander et al.
(1986). Significant volumetric reduction reported in both the frontal cor-
tex (Coffey et al. 1993) and subcortical nuclei (Krishnan et al. 1992) of
depressed patients is further evidence suggestive of structural disruption
in putative frontal-subcortical networks.

Huntington's Chorea and Parkinson's Disease


In Huntington's chorea, Weinberger et al. (1988) and Starkstein et al.
(1988) have demonstrated a strong relationship between caudate atro-
phy on CT scan and cognitive test performance, but not between motor
deficit and caudate atrophy. Further evidence for this finding comes
from two controlled PET studies by Berent et al. (1988) and Maziotta
and colleagues (cited in Cummings 1992) which each separately report
reduced caudate metabolism, correlating highly with cognitive task per-
formance in HC patients but not controls. Hasselbalch et al. (1992) in
an HMPAO SPECT study, also reported correlations between WCST
and caudate rCBF.
In studies of HC and PD patients with and without depression, May-
berg et al. (1992) and Mayberg et al. (1990b) reported reduced meta-
bolism in the orbito-frontal and inferior prefrontal cortex of the de-
pressed HC and PD patients, the extent of which correlated with the
severity of the depression. No such changes were found in the non-
depressed group. Depressed and non-depressed subjects could be identi-
fied on the basis of this pattern of reduction in rCBF, which intriguingly
is almost identical to that reported by Bench et al. (1992) in melancholic
subjects. These studies imply a common pattern of rCBF reduction
across all three disorders when depression is present.

Applying our Understanding of Pathology in


Frontal-subcortical Networks to the Various Clinical
Presentations of Melancholia
This review of the behavioural, cognitive and motor abnormalities of
melancholia, PD and HC highlights the many similarities between these
disorders. Furthermore, the majority of the studies outlined suggest that
dysfunction at the level of the prefrontal cortex and the basal ganglia
contributes significantly to the key clinical expression of melancholic de-
pression, HC and PD. On the basis of the evidence reviewed, it would
Melancholia as a Neurological Disorder 235
seem reasonable to postulate that dysfunction in a number of the func-
tional circuits proposed by Alexander et al. (1986) may explain at least
some of the clinical features of PD and HC and thereby enlighten us
further about the pathogenesis of melancholia.
The variable clinical presentations of melancholia may be understood
by remembering that both functional and structural disruption of the rel-
evant regions or pathways may be responsible. Thus, recurrent episodes
with subsequent recovery may be understood as reflecting intermittent
periods of abnormal function in genetically vulnerable individuals. On
the other hand, the treatment resistance or lack of full recovery often
seen in patients with predominantly late-onset melancholia may be ex-
plained by structural, and thus potentially irreversible, disruption in
these functional networks. The recent MRI studies of elderly melancholic
subjects identifying deep white matter and subcortical grey hyperinten-
sities would support the possibility of such structural lesions. Where the
lesions themselves are of insufficient severity to lead to symptoms, this
might occur with the addition of a stressor such as a negative life event
(Krishnan 1993b).
The added presence of motor agitation in some retarded melancholic
patients may be explained on the basis of variable input from both
dopaminergic and other neurotransmitter systems on the motor loop.
Psychotic symptoms, which are also often present in agitated melan-
cholic subjects, may be related to increased mesolimbic dopamine activ-
ity - analogous to the formulation of Davis et al. (1985), who suggested
this mechanism to explain the co-existence of positive and negative
symptoms in schizophrenia. By acknowledging the potentially variable
and often interdependent involvement of a number of neurotransmitter
systems in the modulation of these functional neural networks involved
in the production of melancholia, we can attempt to encompass the
variations in psychomotor disturbance and psychotic symptoms seen in
individual depressed subjects.

Conclusion
The central hypothesis proposed by ourselves and others (e.g., Krishnan
1993b) - that the clinical triad of depression, bradykinesia and subcor-
tical dementia seen in PD, HC and melancholia is caused by variable
combinations of dysfunction in a number of parallel frontal-subcortical
neural networks - is amenable to testing using a number of different
approaches. We suggest that this hypothesis would best be tested using
236 M.-P. Austin and P. Mitchell
a combination of neuroimaging, biochemical and neurocognitive inves-
tigations in a homogeneous group of melancholies selected by the CORE
measure (Parker et al. 1994). Such a "convergence of evidence" approach
- as advocated by Carpenter et al. (1993) - should clarify this postu-
lated role of the basal ganglia and frontal regions in the pathogenesis of
melancholia.
16
Melancholia and the Ageing Brain
HENRY BRODATY

Phenomenology of Depression in Older Age


If one accepts the converging evidence that some types of functional
depression represent neurological disease (Rogers 1985; Starkstein and
Mayberg 1993), then it stands to reason that the proportion of depres-
sive disorders with psychomotor disturbance (PMD) should be greater
with increasing age, as both cerebrovascular disease, and basal ganglia
shrinkage and vacuolation appear to be linked to both PMD and growing
older (Creasey and Rapoport 1985).
This hypothesis can be tested by comparing the phenomenology of
depression in elderly (usually defined as at least 60 years of age) and
non-elderly patients. Unfortunately, studies have produced inconsistent
results. Older depressives have been described as being more anxious,
preoccupied with physical symptoms, fatigued, withdrawn, retarded, ap-
athetic, inert, uninterested in their surroundings, lacking drive (Zung
1980; Klerman 1983; Ruegg, Zisook and Swedlow 1988), hypochondri-
acal, anxious, self-reproachful and suicidal (American Psychiatric As-
sociation 1968; Kawashimi 1979) and as having more primary insom-
nia (Brown et al. 1984). However, Gurland (1976) reported that the
only significant difference in the phenomenology of older depressives was
their greater frequency of hypochondriacal symptoms. Blazer, Bachar
and Hughes (1987) found no differences in symptoms between middle-
aged and elderly hospitalised melancholic depressives. Musetti et al.
(1989) also argued that they had refuted the stereotype of the agitated,
hypochondriacal elderly depressive, although they acknowledged that
their sample was predominantly derived from outpatients.
Previously, we have argued (Brodaty et al. 1991) that these inconsis-
tencies may be explicable by sampling differences and more importantly

237
238 H. Brodaty

by the failure of researchers (Brown et al. 1984; Musetti et al. 1989;


Burvill et al. 1989) to analyse their data according to sub-categories of
depression.

Empirical CORE Data


We examined the proposition that depression with onset later in life
was phenotypically different to depression occurring before the age of
60 years. In our first study, of 242 consecutive patients with unipolar
DSM-111-defined major depressive episodes (MDE), the 61 elderly patients
had slightly, but significantly, more severe depression (as measured by
the Hamilton [1960] Rating Scale for Depression [HRSD]), substantially
higher Newcastle endogeneity scores (Carney, Roth and Garside 1965)
and significantly greater scores on the CORE I measure. They were
more likely to be delusional and agitated and to have marked loss of
appetite (Brodaty et al. 1991). Psychomotor retardation and age were
not significantly distinguishing factors. In this sample of unipolar de-
pressives, psychosis occurred in no patient under the age of 40 years,
11% of those 40-59 years old and in 33% (N == 20) of the 61 patients 60
years and older. Agitation, as defined by that item in the HRSD, had
a more pronounced gradient, with marked agitation occurring in 4% of
the young, 15% of the middle-aged and 28% of the elderly depressives.
Re-analysis of results after exclusion of the psychotic depressive sub-
group again found an over-representation of agitation in older patients,
indicating that this finding was not merely an effect of psychosis. There
was a trend for CORE scores to increase with age in the non-psychotic
depressives (p = .08), although it should be noted that a weakness of the
CORE I measure was its lack of agitation items and its heavy weighting
to psychomotor retardation.
In our second study (Brodaty et al., in preparation) of an additional
285 consecutive in- and outpatients with unipolar MDE, we focussed
on the interaction of age with the CORE and its sub-scales. The main
differences methodologically from the first study were the use of DSM-III-
R to define MDE and the use of the CORE (from the CORE II studies),
which contained more agitation items as well as clearly defined sub-scales
of agitation, retardation and non-interactiveness.
Once again there was a clear interaction between age and diagnosis,
with psychotic depression occurring in 2.5% of young depressives (3/119
patients aged < 40 years); 6.7% of middle-aged (6/89 patients aged 40-
59 years); and 32.5% of elderly depressives (25/77 patients aged > 60
Melancholia and the Ageing Brain 239
years). Total CORE scores were markedly higher in elderly than non-
elderly depressives (means 15.3 vs. 5.0; F = 115.1, P < 0.0001). CORE
scores for young and middle-aged subjects were similar (see Figure 16.1).
All CORE sub-scale scores were higher in older patients. The most
striking interaction was between age and agitation. Middle-aged depres-
sives were much more agitated than younger patients (F = 17.0, P <
0.001) and the elderly far more agitated than younger and middle-aged
depressives combined (F = 60.9, P < 0.0001). Exclusion of those with
psychotic depression did not alter the findings (see Figure 16.1).
Psychomotor retardation and non-interactiveness received higher rat-
ings in elderly than non-elderly subjects (F = 76.55, P < 0.001 and
F = 36.66, P < 0.001 respectively) even when psychotic depressives
were excluded (P < 0.001 for both analyses). Unlike agitation, there
was no age gradient for retardation or non-interactiveness, as the scores
on these sub-scales for the young and middle-aged patients did not differ
appreciably.
Clearly, psychosis and psychomotor features, especially agitation, oc-
curred more commonly with increasing age in our depressed samples.
While the increased frequency of psychosis in elderly depressives has long
been recognised (Post 1972; Eagles and Whalley 1985; Meyers, Green-
berg and Mei-Tal 1985; Burvill et al. 1989; Conwell et al. 1989), psy-
chomotor phenomena in general and agitation in particular have rarely
been noted.

Psychomotor Changes and Depression in Old Age


Possible explanations for the findings of age-associated psychomotor
changes in depression discussed above will now be considered.

Organic Brain Disease


The findings of age-associated psychomotor changes could merely reflect
the inclusion of a group of older patients with unrecognised organic brain
disease, thereby inflating CORE scores when all elderly depressives are
considered. While there is anecdotal support and some inconsistent
empirical evidence for this proposition in other studies, data do not
support this as a sufficient explanation.
Intuitively, one might expect that age-associated degenerative and
vascular brain pathology would be associated with more organic or neu-
rological types of depression. Indeed, depression appears to be more
240 H. Brodaty

Unipolar major depression


40 20 n

<40 (N = 119)
40-59 (N = 89)
30 15 60 + (N = 77)

a 20 o> 10 -

10 5 -

0 J

Grouping
by age CORE Agitation Retardation Non-inter-
activeness

Non-psychotic unipolar major depression


20 -,

<40 (N = 116)
40-59 (N = 83)
15 - 60 + (N = 52)

o> 10 -

5 -

J
0

CORE Agitation Retardation Non-inter-


activeness

Fig. 16.1. Means on the CORE and Its Sub-scales for Age Groupings. In the
Top Panel all Patients are Included; the Percentage with Psychotic Depres-
sion is Indicated to the Left. In the Bottom Panel Patients with Psychotic
Depression have been Excluded.

common in degenerative conditions such as Alzheimer's disease (Rovner


et al. 1989; Loreck and Folstein 1993), although other reports refute this
(Cummings et al. 1987; Weiner, Edland and Luszczynska 1994). Depres-
sion is also common in Parkinson's disease (Parkinson 1817; Starkstein
and Mayberg 1993) and after cerebrovascular incidents (Ebrahim and
Nouri 1987; Starkstein and Robinson 1993; Burvill 1994). In all of these
Melancholia and the Ageing Brain 241
conditions the putative pathogenesis is disruption of ascending aminer-
gic systems (Zweig et al. 1988; Zubenko et al. 1990; Zubenko and Moossy
1993; Loreck and Folstein 1993; Starkstein and Robinson 1993).
Loreck and Folstein (1993) suggest that some level of cholinergic in-
tegrity is necessary for the expression of depressed mood. In Alzheimer's
disease the cholinergic system appears to be disproportionately affected
(Perry 1986; Zubenko et al. 1990), and it may only be those with suffi-
cient cholinergic activity who are capable of being depressed. This is con-
sistent not only with a report (D'Amato et al. 1987) that patients with
Alzheimer's disease treated with oxotremorine (a long-acting cholinergic
agonist) became depressed, but also with our own similar observation
of patients with Alzheimer's disease treated with tacrine (Brodaty and
Peters, unpublished manuscript).
While depression is common after stroke, its phenomenology has re-
ceived little attention. There is no evidence that depressive psychosis
or psychomotor signs are more common (though psychosis per se is well
described). In Parkinson's disease, affective, anxiety and autonomic fea-
tures have been reported as occurring commonly (Gotham, Brown and
Marsden 1986; Schiffer et al. 1988; Huber et al. 1990; Starkstein and
Mayberg 1993), but not psychosis or agitation. (Retardation, being so
common in Parkinson's disease, would not be helpful in distinguishing
a particular type of associated depression.) In other words, while de-
pression may be more common in these neurological disorders, and these
disorders are more prevalent in late life, these conditions are not neces-
sarily associated with psychosis or psychomotor changes.
The empirical evidence for a distinctive depressive phenotype occur-
ring for the first time in older people requires comparison between early-
onset (EO) and late-onset (LO) depressive disorders. Delusions, guilt,
hypochondriasis and anxiety have been described as more common in
LO depressives by some authors (Post 1962; Pichot et al. 1979; Shin-
fuki 1979; Brown et al. 1984; Meyers et al. 1985; Meyers and Greenberg
1986; Conwell et al. 1989) but as equally common by others (Post 1962;
Hopkinson 1964; Winokur, Behar and Schesser 1980; Greenwald and
Kramer-Ginsberg 1988; Musetti et al. 1989; Nelson et al. 1989). We
found no differences in clinical features between EO and LO elderly de-
pressives in either the CORE I or CORE II samples. In the first sample,
there were no differences in CORE I CORE scores or in individual symp-
toms, though putative aetiological factors such as family history (EO >
LO) and physical ill health (EO < LO) differed in the expected direc-
tion (Brodaty et al. 1991). In the second sample, EO and LO elderly
242 H. Brodaty
depressives' scores on the CORE II CORE and its sub-scales were similar
(Brodaty et al., in preparation).
In conclusion, the first putative explanation that a sub-group of LO
depressives are entirely responsible for the increasing frequency of psy-
chomotor change associated with depression and ageing is supported
neither by examination of depressions occurring in age-associated neu-
rological disorders nor by comparison of the phenomenology in EO and
LO depressives.

Effects of Recurrence and Intervention


A second possible explanation, related to the first, is that the phenotypic
differences result from the greater likelihood of recurrence and concomi-
tant intervention with increasing age. In other words, the chemical and
physiological changes associated with repeated episodes of depression
or its treatment could influence the phenotypic expression of successive
episodes. In both our samples, first episode depressives (N = 81 in the
first sample and N = 106 in the second sample) were compared with
those having a recurrence (N = 161 and N = 179 respectively). They
were not found to differ in regard to symptoms and scores on the CORE
measure or its sub-scales.
Longitudinal phenomenological cohort studies could assist in examin-
ing the influence of recurrence on depressive sub-types, but they have
been lacking until recently. In an important eight-year follow-up study
on the long-term stability of depressive sub-types, Coryell et al. (1994)
reported that psychotic, agitated/retarded and endogenous sub-types
were stable across contiguous episodes in primary unipolar depressives,
but this relationship was weaker between more distant episodes. Of
all the sub-types, the psychotic one was the most consistent, a finding
confirmed by Sands and Harrow (1994), who postulated that psychosis
and affective disorder represented separate interacting factors. It seems
unlikely that recurrence of depression (or its treatment) is exerting a
pathoplastic effect.

Cohort Effects
Patients with depression born before, say 1930, may have been subject
to special influences. Longitudinal study of successive cohorts might
provide a definitive answer to this possibility. We can be reasonably
Melancholia and the Ageing Brain 243
confident, however, that this explanation is improbable, since there ap-
pears to be a gradient effect with age for psychosis and agitation (see
Figure 16.1).

Sodoenvironmental Explanation
A fourth possible explanation is more tenuous. In twelve epidemiolog-
ical surveys reviewed by Wittchen, Knauper and Kessler (1994), the
prevalence of major depression was lowest in the elderly (e.g., Regier
et al. 1993). Henderson (1994) examined and rejected the following pos-
sibilities as sufficient to explain this lower prevalence: sampling bias;
early death of depressed adults; misattribution of depressive symptoms
to physical illness; unwillingness to affirm particular key defining symp-
toms such as lowered mood; and cognitive impairment leading to nega-
tive responses by the surveyed elderly. If the observed low prevalence is
valid, it may be that fewer elderly individuals have sufficient symptoms
to reach case threshold or that the incidence may be the same but the
episodes are briefer. Henderson went on to reject as counter-intuitive the
possibility that the biological basis for depressive disorders is reduced
with ageing. Instead, he favoured a sodoenvironmental explanation cit-
ing evidence that life events become less frequent in late life (Henderson,
Byrne and Duncan-Jones 1981) and that social supports, though fewer
in availability, become more satisfactory. Further, repeated exposure
to adverse experience over a lifetime could gradually increase an indi-
vidual's resistance to depression, so-called psychological immunisation
(Henderson, Montgomery and Williams 1972).
If Henderson's theories hold, and they appear attractive, they are
more likely to apply to non-melancholic depression. In other words,
the lower prevalence may result from removal of non-melancholic sub-
jects from the depressive pool, thereby causing a lowering of the preva-
lence of depression with age and artificially elevating the proportion with
melancholic and psychotic depressive episodes. Unfortunately, such fine-
grained analysis by sub-type is not available from epidemiological data,
although in both of our samples the proportion of unipolar MDEs with-
out melancholia was lower in the elderly (18% in the CORE I and 6.5%
in the CORE II samples) than in the non-elderly (35% and 56.7% re-
spectively). However, when young and middle-aged depressed patients
were compared, patients < 40 years of age had higher rates of non-
melancholic depression (50% and 68% across the two samples) than those
aged 40-59 years (42% in each sample). These data, which run counter
244 H. Brodaty

to the age gradient for psychosis and agitation, do not necessarily exclude
this fourth explanation, as an age threshold effect may be in operation.
In summary, the differential removal of non-melancholic elderly de-
pressives from the clinical or epidemiological samples may well be oc-
curring through a process of maturation of resilience to adversity, but it
is unclear whether this is of sufficient magnitude to explain the higher
rates of depression with psychomotor features and psychosis in later life.

Age-associated Neuropathology
In order to examine the fifth proposition, that age-associated brain
changes may be the cause of the phenotypic differences in depression
in late life, it is necessary to consider what normally occurs with ageing.

The Brain in Normal Ageing


Consistent histological findings from autopsy studies include age-related
reduction in brain weight, brain volume, cortical volume, regional corti-
cal neuronal numbers and ventricular dilatation. These changes are not
uniform, being more prominent in the white than the grey matter and in
the hippocampi, especially the subiculum, than other areas (Creasey and
Rapoport 1985; Esiri 1994). Ventricular dilatation and the variance in
ventricular size appear to increase with age too (Creasey and Rapoport
1985; Coffey et al. 1992). Microscopically, there is definite but modest
cell loss although, more typically, there is shrinkage of nerve cells. This
occurs mainly in large neurons with a resulting relative increase in the
numbers of small neurons and glia, particularly in the mid-frontal and
superior temporal cortices (Davis, Mirra and Alazraki 1994). Lipofuscin
accumulates in some neurons, nucleolar volume shrinks and dendrites
decrease in number and length, as well as in their numbers of spines and
synapses (Esiri 1994). With age, there is also a tendency for neurofib-
rillary tangles and senile plaques to form, vascular amyloid to deposit,
atheromatous plaque to accumulate, arteriolar walls to thicken and cere-
bral blood flow to decrease.
Post-mortem neuropathological brain studies unfortunately are sub-
ject to selection bias, technical and fixation artifacts, pre-morbid ill-
ness, and the effects of whatever caused death (Creasey and Rapoport
1985). Imaging studies can overcome many of these problems (Cof-
fey et al. 1992). Those undertaken in normal older subjects have been
usefully summarised by Beats and Levy (1992), Baldwin (1993), and
Krishnan (1993b). Consistent findings on computed tomography (CT)
Melancholia and the Ageing Brain 245
scanning are progressive cortical atrophy, increasing size of lateral ventri-
cles, greater CSF volume and greater frequency of hypodensities. Most
of these changes become more noticeable after the age of 50 or 60 years.
Magnetic resonance imaging (MRI) confirms, and provides greater de-
tail of, CT evidence of atrophy and areas of hypodensity. Coffey et al.
(1992) found, in a study of 76 healthy adults (age range 30-91 years),
that the volumes of the cerebral hemispheres declined by 0.23% per year,
the frontal lobe by 0.55% per year, the temporal lobe by 0.28% per year
and the amygdala-hippocampal complex by 0.30% per year, consistent
with autopsy results. Ventricular volumes increased at a rate of 2.8%
per annum for the third ventricle and 3.2% per year for the lateral ven-
tricles. Subcortical hyperintensities of deep white matter increased by
6.3% per year. Decreases with age in the volumes of the caudate and
lenticulate nuclei, put amen and corpus callosum appear to be out of pro-
portion to general brain shrinkage (Krishnan et al. 1992; Murphy et al.
1992). Fazekas (1989) also reported that subcortical grey matter and
deep white matter changes occurred much more commonly with age (in
11% of subjects aged 31-40 compared to 83% of those > 70 years of
age), and with cerebrovascular risk factors. There is also a much higher
prevalence with age of leukoencephalopathy (George et al. 1986) - ar-
eas of signal hyperintensities on T2-weighted images, primarily found
scattered in the periventricular white matter, deep white matter, basal
ganglia and pons. While these are common, confluent and large lesions
appear to be unusual or rare (Krishnan 1993b).

White matter hyperintensities identified by MRI have been the fo-


cus of much investigation. They have been observed in older persons
without neurological deficits as well as those with dementia. Method-
ological difficulties have impeded understanding of the neuropathologi-
cal changes underlying these white matter hyperintensities. Associations
have been reported between white matter hyperintensities and myelin
pallor, demyelination, diffuse vacuolation, gliosis, dilated peri vascular
spaces, vascular ectasia, lacunar infarction and reduction of glial cell
density (Awad, Johnson and Spetzler 1986; Kirkpatrick and Hayman
1987; Marshall et al. 1988; Davis et al. 1994). Chronic vascular insuf-
ficiency and leakage of serum proteins from cerebral vessels may play a
role in pathogenesis (Davis et al. 1994). Larger focal hyperintensities ap-
pear to be true infarcts (Marshall et al. 1988). Hyperintensities anterior
to the frontal horns of the ventricles have been reported to increase with
age in most healthy volunteers and to correlate with increased intersti-
246 H. Brodaty

tial fluid secondary to patchy loss of ependyma with astrocytic gliosis


(Sze et al. 1981).
These MRI changes appear to have an effect on cognitive functioning,
though there may be a critical site or volume threshold before such effects
can be demonstrated. Deep white matter hyperintensities, especially if
marked, have been reported to correlate with slower speed of mental
processing and poor attention span (Ylikoski et al. 1993), poorer visual
memory (Kasahara et al. 1993), and poorer neuropsychological func-
tion in general (Matsubayashi et al. 1992). However, Almkvist et al.
1992 found no relationship between neuropsychological measurements
and white matter hyperintensities in healthy aged adults, and confirmed
that the degree of white matter hyperintensities was related to age and
blood pressure (Drayer 1988).

The Brain in Depressed Older Patients


Proof that age-associated pathology plays an aetiological role in depres-
sion must demonstrate either an additional but interacting factor such
as genetic predisposition, or additional pathology, that is, beyond that
expected in "normal" ageing. Regarding the former, while most studies
report lower familial loadings for late-onset depression than for early-
onset depression (Hopkinson 1964; Brodaty et al. 1991), comparisons of
late-onset depressives against non-depressive age-matched controls have
been lacking. The other possibility, that there is excess pathology in
the older depressives, has attracted more inquiry, principally through
imaging studies.

Imaging Studies in Depression


CT Studies in Depression. Jacoby and Levy (1980), in their seminal
study on CT findings in elderly depressives, reported that there was
a sub-group of eight patients with increased ventricles to brain ratios
(VBRs) who were older, had late-onset depression and were more likely
to have an endogenous type of depression. At follow-up they had a higher
mortality rate, mainly from cardiovascular and cerebrovascular disease
(Jacoby, Levy and Bird 1981). This association between poor prognosis
of depression, ventriculomegaly and cerebral hypodensity has since been
confirmed in a number of studies (Pearlson and Veroff 1981; Targum
et al. 1983; Shima et al. 1984; Kellner, Rubinow and Post 1986; Schlegel
and Kretzschmar 1987; Abas, Sahakian and Levy 1990). Generally,
depressed patients with abnormal CT findings have been older and have
Melancholia and the Ageing Bradn 247
had poorer scores on cognitive testing. Methodological limitations of
these studies, which are well described by Macdonald (1992) and by
Baldwin (1993), include small sample sizes, selective patient referrals
(e.g., those receiving ECT), older age of late-onset depressive disorders,
and non-uniform classification and rating of lesions.

MRI Studies in Depression. In a number of MRI studies, large deep


white matter hyperintensities and basal ganglia lesions stand out as
occurring more frequently than expected in older patients with depres-
sion. Coffey et al. (1988) and Coffey et al. (1989) found that grade two
and grade three periventricular hyperintensities, as well as grade two
and grade three deep white matter intensities were more common in
depressed patients than in controls. Lesions of the basal ganglia were
present in 51% of depressed patients as against 5% of controls.
Similarly, Figiel et al. (1991) found a strikingly high prevalence of
basal ganglia lesions in late-onset depressed patients (60%) compared to
those patients over 60 with early-onset depression (1%). The two groups
did not differ on overall occurrence of deep white matter lesions; how-
ever, the late-onset group had more deep white matter hyperintensities
larger than 1 cm in diameter. Ninety percent of the late-onset group
had either a basal ganglia lesion or a lesion greater than 1 cm in the
deep white matter, primarily in the frontal lobe. Zubenko et al. (1990)
found that depressed patients with cognitive impairment had increased
cortical atrophy compared to those without cognitive impairment, and
that depressed patients in general had more cortical atrophy, lacunar
infarction and leukoencephalopathy than normal controls. Husain et al.
(1991) demonstrated that depressed patients, both young and old, had
smaller putamen and caudate nuclei than controls (as quoted in Krish-
nan 1993b).
Attempts have been made to understand the neuroanatomical sub-
strate by linking these findings with the well-known cognitive impair-
ment found in depression. Four out of five MRI studies of elderly de-
pressives reviewed by Baldwin (1993) found a relationship between white
matter changes such as periventricular hyperintensities or subcortical
white matter lesions and poorer cognitive function, even after correc-
tion for education. More studies than not have reported that, when age
of onset of depression is used to define depressive sub-groups, late-onset
elderly depressives are more likely to have white matter hyperintensities.
Certain of the MRI lesions seem specific to depression in later life
rather than a non-specific finding associated with psychiatric disorder
248 H. Brodaty

in general. In a study of 229 psychiatric subjects with various diagnoses


and ages and 154 normal volunteers (Lesser, Miller and Boone 1991),
the only significant difference found was a greater severity of deep white
matter hyperintensity signals in subjects over 45 years old with major
depression. There was no greater prevalence of hyperintensity signals in
bipolar disorder. Also the presence of hyperintensity signals per se did
not correlate with diagnosis; it was only when severity of the signals was
considered that the association with major depression was found. Lesser
et al. (1991) concluded that older age is an indirect risk factor for the
presence of white matter hyperintensity signals, probably as a result of
vascular changes associated with concurrent medical illnesses.
Other evidence indicates that the MRI lesions and the depression are
both consequent to cerebrovascular disease. Associations of MRI lesions
with vascular risk factors including hypertension, history of ischaemic
heart disease and diabetes mellitus have been reported by some (Coffey
et al. 1989; Lesser et al. 1991) but not others (Harrell et al. 1991; Rabins,
Pearlson and Aylward 1991).
Could the depression be a harbinger of early dementia? This is con-
sidered shortly, but here we note that when subjects with dementia as
well as normal controls are compared with depressed patients, the degree
of MRI brain atrophy for the depressed group is intermediate between
that found for the other two (Rabins et al. 1991).

Functional Imaging Studies in Depression. Morphological imaging


studies lead to speculation about possible relationships between white
matter hyperintensities, clinical characteristics of depression, risk fac-
tors for depression and sub-types of depression (Baldwin, 1993). These
speculations remain unresolved; unfortunately, functional imaging has
not been able to resolve them either. The few studies reported, us-
ing single photon emission computed tomography (SPECT) or positron
emission tomography (PET), have been in younger patients. The con-
sistent finding across these studies is of decreased left frontal metabolic
rate localised to the dorsal anterior lateral prefrontal cortex and related
to the severity of the depressive symptoms in all groups. On activation,
there appears to be hypofrontality and decreased basal ganglia metabolic
rate (Beats and Levy 1992).

Summary of Imaging Studies on Depression in Late Life. Clearly,


the changes in the brains of older depressed patients, principally deep
Melancholia and the Ageing Brain 249
white matter hyperintensities larger than lcm and basal ganglia le-
sions, are greater than would be expected from normal ageing. Krishnan
(1993b) concluded that depressed patients have increased frequency of
lesions in the frontal lobe and basal ganglia as well as smaller basal
ganglia, suggesting that these structures are involved in the pathophys-
iology of depression. They supported findings by Starkstein and Robin-
son (1993), who suggested that in post-stroke patients lesions of the
caudate nucleus and the frontal cortex may play a role in the develop-
ment of post-stroke depression. Krishnan (1993b) postulated that the
frontal lobes and caudate nucleus are the neuroanatomical substrates of
depression. Further evidence was provided by the high prevalence of de-
pression in basal ganglia disorders such as Parkinson's and Huntington's
diseases and the diminished metabolic glucose rates in the basal ganglia
of depressed patients studied using PET.
It is of interest that younger depressed patients may also have white
matter hyperintensities. Dupont et al. (1990) reported that almost half
of a group of patients with bipolar disorder (mean age mid-30s) had
white matter hyperintensities and these were associated with a high
number of previous admissions and poorer cognitive test scores. Thus,
although age is one important risk factor for white matter hyperinten-
sities in depression, it is not the only one (Baldwin 1993).

Pathological Studies in Depression


In 1883, Mairet conceptualised "melancholic dementia," observing that
in a series of patients with both organic brain changes and reversibil-
ity, there were temporal lobe changes at autopsy (quoted in Emery and
Oxman 1992). Later studies (Newton 1948; Roth 1955) reported de-
generative changes in post-mortem examinations of depressed patients
who did not go on to develop dementia. In a recent autopsy study of
late life depression, Bowen et al. (1989) reported that four of seven older
patients with depression showed sulcal widening, ventricular dilation,
white matter loss and microscopic vascular changes.

Follow-up Studies
If cerebrovascular structural lesions are the basis for melancholic de-
pression occurring later in life, then one would expect to find evidence
of progression on follow-up. Dementia, however, has not been found
to be clearly more common in elderly depressives (Kay 1962; Murphy
1983; Kiloh, Andrews and Neilson 1988; Lee and Murray 1988; Mur-
phy et al. 1988; Burvill et al. 1991). On the other hand, retrospective
250 H. Brodaty
(Agbayewa 1986; Burns, Jacoby and Levy 1990) and case control data
(Jorm et al. 1991) suggest that a history of depression is a risk factor
for Alzheimer's disease (and not merely its prodrome). Possible mecha-
nisms include effects of recurrent or prolonged catacholamine depletion,
hypercortisolaemia, effects of treatment, degeneration inherent in the
depressive disease process or an independent third factor (Emery and
Oxman 1992).
There is greater than expected mortality in depressed older inpatients
(Stendstedt 1959; Kay 1962; Post 1972; Jacoby et al. 1981; Murphy
1983; Baldwin and Jolley 1986; Murphy et al. 1988; Burvill et al. 1991)
and community subjects (Kay and Bergmann 1966; Copeland 1988).
Suicide accounts for only a small proportion of these deaths: 2% in the
study from Murphy et al. (1988) and 12% in the study from Burvill
et al. (1991). The majority of deaths are from non-cerebral causes. In
their follow-up study, Burvill et al. (1991) reported that deaths resulted
from cardiovascular disease in 27% of subjects, malignancy in 21%, res-
piratory disease in 15% and cerebrovascular and renal diseases in 9%
each. Older age, male sex, and chronic physical ill health at index as-
sessment, but not severity of depression, type of depression, endogeneity
or age of onset of depression, increased the likelihood of death (Burvill
and Hall 1994). Cognitive impairment and CT scan abnormalities have
been found by some but not others (Burvill and Hall 1994) to presage a
poorer prognosis.
The explanation for why the depressed elderly die earlier is not clear.
The association of death and depression remains, even after account-
ing for physical health problems (Bruce and Leaf 1989). Poor self-care,
including inadequate nutrition and even the "wish to die" may play a
part. Treatment probably has a protective role. Depression-related bi-
ological changes such as hypothalamic-pituitary-adrenal axis activation
and compromised immune function are theoretically appealing but have
little supportive evidence (O'Brien and Ames 1994). The possibility that
depression could be a risk factor for another disorder such as dementia
which, in turn, causes increased mortality, seems unlikely, as most of
the increase in mortality occurs in the first one or two years after the
index episode (Murphy et al. 1988; Jorm et al. 1991). These findings are
qualified by the failure to undertake fine-grained analyses such as look-
ing at sub-types of depression and dementia. An intriguing exception is
the report by Burvill and Hall (1994) of a lower mortality in patients
with psychotic depression compared to those with other types of depres-
sion. Another possibility might be that people with depression destined
Melancholia and the Ageing Brain 251
to develop vascular dementia die earlier from cardiovascular disease (as
reported by Murphy et al. 1988; Burvill et al. 1991), thereby obscuring
this possible relationship.

Conclusion and Hypothesis


A putative anatomical localisation for melancholic depression has been
proposed (Austin and Mitchell 1995; and Chapter 15). Recurrent epi-
sodes with subsequent recovery are understood, in these models, as re-
flecting intermittent periods of abnormal function, possibly precipitated
by life events, in genetically vulnerable individuals. Late-onset melan-
cholia may be explained by structural and thus potentially irreversible
disruption in these functional networks. Where the lesions themselves
are of insufficient severity to lead to symptoms, the addition of a stressor
such as a negative life event might be required for depression to occur
(Krishnan 1993b). Austin and Mitchell (1995) conceded that, while the
retardation in melancholia is thus explained, the motor agitation is more
difficult to incorporate into this hypothesis. They suggested that a para-
doxical increase in dopamine function in these same pathways or a more
extensive dysfunction in the motor loop may be occurring. We have
confirmed, in samples of referred patients with major depression, that
psychomotor phenomena, particularly agitation, as well as psychosis,
occur more commonly with age. Our working conclusion is that struc-
tural brain pathology underlies these phenomena in older melancholies.
Such pathology may be a factor separate from, but interacting with,
depression. In other words, both unipolar major depression and ageing
combine to increase the risk of psychosis and psychomotor change, but
either acting independently is not likely to produce these phenomena.
17
Magnetic Resonance Imaging in Primary
and Secondary Depression
IAN HICKIE
CATHERINE HICKIE
ELIZABETH SCOTT
KAY WILHELM

Structural Imaging in Neuropsychiatry


Given the capacity of magnetic resonance imaging (MRI) for

(i) precise visual presentation of neuroanatomical structures;


(ii) differentiation of grey and white matter structures;
(iii) detection of white matter lesions;
(iv) reconstruction of images in multiple planes; and
(v) the utilisation of differing scan sequences to highlight varying tissue
characteristics
it is well suited for investigating brain structure in patients with neu-
ropsychiatric disorders (Andreasen 1988; Krishnan 1993b). Addition-
ally, as it does not involve ionizing radiation, repeated examinations are
possible. This is of key importance in unravelling diagnostic issues in
psychiatry. Specifically, for those interested in the relationships between
the affective and degenerative disorders of late life, the combination of
MRI and longitudinal clinical methodologies is essential.
For patients presenting with psychomotor change, cognitive dysfunc-
tion and/or psychotic features, MRI provides ready visualisation of the
limbic system and other key basal ganglia structures which have been
proposed as providing critical neuroanatomical pathways (Krishnan 1991;
Krishnan 1993b; Drevets et al. 1992; Cummings 1993; Mayberg 1993).
As MRI technologies are increasingly integrated with other functional
scanning systems (positron emission tomography [PET] and/or single
photon emission computed tomography [SPECT]), the co-registration of
data provides the capacity to describe precisely the anatomical regions
exhibiting differences in regional cerebral blood flow (rCBF) or ligand
binding in neuropharmacological and neurochemical receptor studies.

252
Magnetic Resonance Imaging 253
Relationships between neuroanatomical lesions and their consequent lo-
cal, and broader regional, biochemical effects can now be directly ex-
plored. A range of other MRI applications under development, includ-
ing functional MRI scanning and MRI spectroscopy (though beyond
the scope of this chapter), will provide further evidence of the highly
localized correlates of specific brain activities and the neurochemical
characteristics of tissue in key regions.
Earlier neuropathological and computed tomography studies of cere-
bral structures in patients with affective disorders suggested reductions
in global, and certain regional, cerebral volumes in older patients (see
Krishnan 1993b). Results across studies presumably represent variabil-
ity in the range of depressive disorders and age groups under examina-
tion. With MRI it is now possible to evaluate specific brain volumes in
vivo (see Coffey et al. 1993) and also to evaluate directly the likelihood
that disruptions to proposed frontal-subcortical-basal functional circuits
(Alexander, DeLong and Strick 1986; Alexander, Crutcher and DeLong
1990) actually account for the specific behavioural phenomena (mood
change, psychomotor dysfunction, cognitive impairment and psychosis)
of interest (see Krishnan 1993b; Carroll 1991). As these phenomena also
occur across a variety of other psychiatric and neurological disorders, the
neuroanatomical and clinical correlates of such phenomena can also be
studied in those settings (see Table 17.1). To date, MRI studies in de-
pression have focused largely on the disorders of late-life, where regional
brain changes (e.g., of the subcortical white matter and basal ganglia)
have been replicated. Investigations of younger patients with bipolar
disorders have suggested that similar white matter abnormalities may
be occurring, but such studies await replication and, more importantly,
longitudinal re-evaluation to exclude the possibility of other emergent
CNS disorders (notably multiple sclerosis [MS]).
Although it is unlikely that the whole spectrum of affective disorders
can be accounted for simply by fronto-striatal neuroanatomical mod-
els (see Chapters 3 and 15), advances in the understanding of late-life
depression and depression in the setting of other neurological disorders
(see Starkstein and Robinson 1993) are of specific relevance to melan-
cholia research. Importantly, such developments depend not only on
the development of sophisticated imaging technologies but also on de-
tailed behavioural measurements, careful hypothesis-driven cohort selec-
tion and comparisons with well-characterised neurological control groups
(see Chapter 3 and Table 17.1). Various neuroanatomical and neuro-
chemical models which have been developed (see Chapter 3) do provide
Table 17.1. Clinical features compared in melancholia and relevant neurological and neurodegenerative disorders

Parkinson's Huntington's Multiple Multi-infarct Alzheimer's


Melancholia Disease Disease Sclerosis Dementia Disease
Depressed mood 100% 20-40% 25-45% 30-55% 20-30% 10-20%
Cognitive changes:
- subcortical
- cortical

Psychomotor features:
- non-interactiveness
- retardation
- agitation
- other Rigidity Chorea Weakness Gait disturbance Gait disturbance
and tremor Fatigue
Incoordination

Manic episodes 10-20% < 10% > 10% > 10% < 10%

Psychotic features
Depression responds
to TCA/ECT
Extrapolations with regard to melancholia are based upon literature examining patients with severe disorders, typically of
late onset.
Table 17.2. Neurobiological correlates compared in melancholia and relevant neurological and
neuro degenerative disorders.

Parkinson's Huntington's Multiple Multi-infarct Alzheimer's


Melancholia Disease Disease Sclerosis Dementia Disease

MRI scanning
- cortical atrophy 4-4- 4- + — 4-+ 4-4-4-
- frontal atrophy 4-4-4- 4- -f — +4- 4-+
- temporal atrophy 4-4- 4- + — 4-+ 4-4-
- parietal atrophy -f 4- + - 4-+ 4-4-4-
- white matter lesions 4-4-4- -f 4- 4-4-4- 4-4-4- 4-
- reduced basal ganglia volume 4-4-' ++4- 4-4-+ 4-+ 4-

rCBF (PET/SPECT)
- global 4- 4- 4 - 4 - 4-4- 4-4-
- frontal 4-4- 4- 4-4- 4-4- 4-4- 4-4-
- temporal 4-4- 4- 4- 4-4- 4-4- 4-4-
- basal ganglia 4-4- 4-4-4- 4-4-4- 4- 4-4- 4-4-
- parietal 4- + 4- 4- 4-4- 4-+4-
Extrapolations with regard to melancholia are based upon literature examining patients with severe disorders, typically of
late onset.
Possible pathophysiology and risk factors compared in melancholia and relevant neurological and
erative disorders

Parkinson's Huntington's Multiple Multi-infarct Alzhe


Melancholia Disease Disease Sclerosis Dementia Dise
athophysiology
sites
nglia ++
al white matter ++
+
cal involvement
amines ++
e -f+
line +
++
nd laboratory risk factors
l vascular disease +
n +
lele) + - - - - +
ory
on -f-h -f + — + -f
a + + +++ — + +
olipid antibodies + — — — + +
ons with regard to melancholia are based upon literature examining patients with severe disorders, typicall
Magnetic Resonance Imaging 257
plausible explanations for severe depressive disorders and/or melancho-
lia, and depressions in the context of structural brain changes. They
also suggest future directions for the investigation of aetiological factors
in the melancholic and psychotic depressive disorders (Krishnan 1993b;
Cummings 1993; Mayberg 1993).

MRI Scanning in Late-onset depression


The large body of work by researchers at Duke University (see Krishnan
1993b and Coffey et al. 1993) has led to the development of a specific
model for late-onset (age > 50 years) depressive disorders. Initial MRI
scanning studies (Krishnan et al. 1988) in elderly patients presenting for
ECT treatment suggested that regions of hyperintensity in the subcorti-
cal white matter and basal ganglia could not be accounted for simply by
age and, contrary to prevailing opinion, that such lesions are of clinical
significance. Given their location within key fronto-striatal circuits, the
lesions were also proposed as having aetiological significance (Krishnan
1993b). Subsequent replication studies have confirmed the increased
prevalence (30-80%) of such lesions in elderly patients with depression
and have extended understanding of the clinical and neurocognitive sig-
nificance of such abnormalities (Brown et al. 1992; Coffey et al. 1993;
Krishnan et al. 1993; Hickie et al. 1995). Patients with depression ap-
pear to have an increased prevalence of anterior (frontal) lesions (while
incidence of such lesions does not appear to be increased in other neu-
ropsychiatric disorders). For example, Brown et al. (1992) demonstrated
in a sample of over 200 patients that, when a range of neuropsychiatric
disorders was studied, extensive white matter lesions were only clearly
increased in patients with "major depression" and, within that group,
were increased only in older (> 45 years) patients.
Computed tomography and MRI studies have suggested a wide range
of other non-specific brain abnormalities including reduced cerebral vol-
ume, reduced cerebellar volume, increased pituitary size and ventricular
enlargement in older patients with depression (Krishnan 1993b). Impor-
tantly, frontal lobe volumes appear to be particularly reduced (Coffey
et al. 1993). This range of abnormalities suggests both global and specific
regional neuronal loss in older depressed patients, though such changes
exceed those predicted by normal ageing (see Chapter 16). Dolan et al.
(1990) reported increased Ti-values on MRI in the frontal white matter
of unipolar depressives compared with controls, suggesting functional
as well as structural frontal lobe abnormalities in these patients. Ad-
258 /. Hickie, C. Hichie, E. Scott and K. Wilhelm
ditionally, Krishnan et al.'s (1990; 1992) description of specific reduc-
tions in caudate and putamen nuclei volumes are of great interest to
those interested in the cognitive and motor abnormalities accompanying
melancholia (see Chapters 3, 9 and 15).
While the description of such specific neuroanatomical changes in
older depressed patients has transformed aetiological and clinical theo-
rising in geriatric psychiatry, it is important to view such developments
in context. Most studies have been performed within highly selected,
elderly patients who have been hospitalised for physical treatments (in-
cluding ECT) and/or for concurrent medical assessment. Clinically, such
patients have severe disorders, both phenomenologically (with high rates
of psychomotor change, cognitive impairment and/or psychotic features)
and in terms of functional disability. The cohorts described demon-
strate high rates of known risk factors to cerebrovascular disease (notably
hypertension), but tend not to have family histories of classical early-
onset affective disorders (Hickie et al. 1995). The comparability of these
patients with those described by Alexopoulos et al. (1993a) as having
"depression with reversible dementia" (i.e., characterised by depression
and severe, though reversible, cognitive impairment at presentation, and
poor longitudinal course) remains to be clarified.
The power of integrating MRI with functional imaging (SPECT) is
well demonstrated by the study of Lesser et al. (1994), who investigated
regional cerebral blood flow (rCBF) in older drug-free depressed pa-
tients. They reported global reduction in rCBF, with the orbito-frontal
and inferior temporal areas being affected bilaterally and preferential
reduction in rCBF in the right hemisphere. They also noted, however,
that "there appeared to be a subgroup of patients who demonstrated
large areas of white matter hyperintensity (on MRI) and low rCBF."
Such studies indicate the likely functional significance of such structural
changes in patients with primary depressive disorders.

Clinical Significance of MRI Changes in Severe Depression


The short-term clinical significance of subcortical white matter and basal
ganglia changes in older depressed patients has received some atten-
tion. The Duke University group has demonstrated that the presence of
such abnormalities predisposes to increased side effects, notably delir-
ium, when such patients receive psychotropic medication or ECT (Figiel
et al. 1989; Figiel et al. 1990a; Figiel, Krishnan and Doraiswamy 1990b).
Such lesions are clearly associated with demonstrable cognitive impair-
Magnetic Resonance Imaging 259
ment, and this cognitive impairment may not remit despite effective
treatment of the depression. Patients referred to tertiary-referral affec-
tive disorders services demonstrate very high rates of such MRI abnor-
malities, suggesting that the presence of such lesions has contributed to
prior poor response to conventional therapies (Hickie et al. 1995).
To date, little other longitudinal evidence is available. Coffey et al.
(1991) performed repeat MRI scanning on 35 patients who had been pre-
viously scanned and then had received ECT treatment. Although the
focus of the study was on the potential to detect cerebral changes fol-
lowing ECT, the authors did note that five patients had more extensive
white matter lesions on re-evaluation. They interpreted such differences
to be indicative of progression of an underlying (probably cerebrovascu-
lar) disease process. Ananth et al. (1993) re-scanned 16 healthy older
adults after 3 years and found little progression in white matter hyperin-
tensities (WMHS), though ventricle-to-brain ratios (VBR) did increase.
In two of the patients, large increases in both WMHS and VBR were
noted and were associated with uncontrolled hypertension in one and
the onset of a major depressive episode in the other.
Alexopoulos et al. (1985, 1987, 1988, 1993a) and Alexopoulos, Young
and Meyers (1993b) have performed a series of clinical studies examin-
ing various aspects of depressions with severe (and initially reversible)
cognitive impairment. They concluded (1993b) that "Biological changes
associated with dementing disorders may be the pathogenic background
in a considerable number of late-onset depressives who have not yet de-
veloped irreversible dementia." The implication of these studies is that
the clinical syndrome of depression in late life with concurrent severe
cognitive impairment is largely a variant of the more prevalent demen-
tias. Longitudinal MRI and clinical studies are essential to evaluate this
specific hypothesis.

MRI Changes and Longitudinal Course: Studies at the MDU


At our MDU we examined the short-term predictive significance of ex-
tensive white matter lesions in a highly selected group of 39 hospitalised
patients with severe depressive disorders (inclusion criteria were age
> 60, presence of psychotic features, prior treatment resistance and/or
having risk factors to, or previous symptoms suggesting, cerebrovascular
disease [Hickie et al. 1995]). The patients had a mean age of 64.4 years,
and 79% had a history of or current psychotic symptoms. Thirty-two
of the patients had unipolar depressive disorders, while seven were in
260 /. Hickie, C. Hickie, E. Scott and K. Wilhelm
the depressed phase of a bipolar disorder. Thirty-five of 39 had clinical
diagnoses of melancholia, while the mean CORE score (from CORE II)
of the groups was 18.7 (SD = 9.7), indicating that the group was char-
acterised by severe psychomotor disturbance. Sixty-seven percent of the
sample had received ECT treatment previously, while 49% had either a
history of hypertension or symptoms suggesting cerebrovascular disease.
White matter hyperintensities on MRI were (i) strongly correlated
with age (r = 0.70, P < 0.001); (ii) associated with the absence of a
family history of affective disorder (P = 0.002); (iii) associated with the
onset of first affective episode after the age of 50 years (P = 0.03); and
(iv) tended to be associated with a history of hypertension (P = 0.08).
They were associated with our key clinical interest, namely, impaired
psychomotor speed (in this instance measured electronically, r = 0.22-
0.54, depending on specific tasks [see Chapter 9]).
Although the sub-group was small (N = 8), among non-psychotic
patients WMHS were strongly correlated with the CORE and its sub-
components (for the CORE r = 0.78, P = 0.02; for non-interactiveness
r = 0.61, P = 0.11; for retardation r = 0.86, P = 0.007; for agitation
r = 0.70, P = 0.05). Among psychotic patients small non-significant
negative correlations between CORE and WMHS were noted. As noted
previously (see Chapters 3 and 12), patients with melancholia who also
have psychotic features present clinically differing forms of psychomotor
change (increased non-interactiveness and agitation), respond preferen-
tially to ECT (see Chapter 10) and, consequently, are likely to have
differing pre-disposing neurobiological risk factors. This initial study
sample was strongly weighted towards psychotic patients, and the appar-
ent relationship between psychomotor disturbance and WMHS detected
here awaits replication in a larger, more representative sample.
The extensiveness of WMHS on MRI was correlated inversely with the
clinician's ratings of global response to conventional treatment strate-
gies (r = — 0.44, P < 0.01; i.e., WMHS predicted a poorer treatment
response). Such effects were true of patients who received pharmacother-
apy alone (r = —0.49, P < 0.05) and those who received electroconvul-
sive therapy after pharmacotherapy (r = — 0.42, P = 0.06). Although
WMHS were strongly correlated with age (r = 0.70, P < 0.001), age
per se was only weakly predictive of outcome (r = —0.32, P = 0.05). In
hierarchical regression analyses, age (entered first) accounted for 10% of
the variance in outcome, while the extent of WMHS (entered second)
accounted for a further 9% of the variance. That is, the poor treatment
Magnetic Resonance Imaging 261
response of these patients did not appear to be a simple function of
ageing.
The long-term clinical significance of subcortical and basal ganglia
changes on MRI remains to be established. Our own clinical re-exam-
ination of 37 of the 39 patients at a mean of 14.1 months (range 6
months to 2 years) after initial scanning indicated high levels of resid-
ual disability attributable to the combination of advanced age, residual
depression and residual cognitive impairment. Fifty-nine percent were
rated by the research psychiatrist as at least moderately improved in
terms of depressive symptoms and/or functioning, leaving 22% with only
mild improvement and 19% who were not improved or worse. Of the
37 patients re-assessed, 12 (32%) were living in institutional care (e.g.,
nursing home, long-term psychiatric care). Three patients had died,
including one suicide. Although no patient clearly met DSM-III-R crite-
ria for dementia, 10 patients were rated as having "probable" dementia
syndromes. Not surprisingly, given that initial selection criteria had
included a history of hypertension or cerebrovascular symptoms, those
with probable dementia were all considered likely to be vascular in ori-
gin (though of variable causes, including one patient with documented
anti-phospholipid antibody syndrome).
The baseline MRI ratings (subcortical white matter and grey nuclei
hyperintensities) and a range of clinical variables (age, age of onset
of affective disorder, duration of depressive episode, history suggestive
of cerebrovascular disease, history of hypertension, psychotic features,
bipolar disorder and treatment with ECT for the index episode) were
assessed as predictors of long-term outcome. The outcome variable
was the longitudinal course of illness since discharge (LC) (rated 1-
8, with, for example, 1 indicating "recovered and continuously well," 4
indicating "partial recovery" and 8 indicating "worse," including sui-
cide). Older age was the strongest predictor of a poor longitudinal
course (r = 0.49, P = 0.002), while older age at onset of affective dis-
order had a similar predictive value (r = 0.43, P = 0.007). Subcortical
WMHS (r = 0.40, P = 0.014), but not grey nuclei changes (r = -0.14),
also predicted a poor outcome. A history of cerebrovascular disease
had some predictive value (r = 0.34, P = 0.04). In a stepwise multi-
ple regression analysis with LC as the dependent variable, and age and
WMHS as independent variables, age alone emerged as a significant pre-
dictor, explaining 20% of the variance (F = 8.59, P = 0.006). For global
clinical rating of final outcome (scored 1-5, with 1 indicating worsened
symptoms or function and 5 indicating complete recovery), subcortical
262 /. Hickie, C. Hickie, E. Scott and K. Wilhelm

WMHS did have some predictive capacity (r = -0.31, P = 0.06), while


age (r = -0.21), age at onset of affective disorder (r = -.20), history
of cerebrovascular disease (r = -0.08) and grey nuclei hyperintensities
were not useful predictors.
In regard to the development of the probable dementia syndromes,
older age at onset of affective disorder (61.3 vs. 44.3 years, t = 2.57, P =
0.015), and more extensive subcortical WMHS (scored 0-6, mean 3.60
vs. 1.93, t = 2.99, P = 0.005) predicted cognitive decline. The remain-
ing clinical and historical variables did not predict later cognitive im-
pairment. A larger patient cohort is now being followed longitudinally
(clinically and by repeat MRI scanning) to extend these initial findings.
In summary, our own initial study of MRI in highly selected pa-
tients with severe depressive disorders indicated that a high proportion
(44%) of such patients had extensive subcortical WMHS (presumably
secondary to hypoperfusion and neuronal loss). We concluded that such
changes were clinically significant, as they correlated with demonstra-
ble neuropsychological (''subcortical") impairment, poorer response to
standard therapies and poorer longitudinal course. A significant sub-
proportion of these patients appear to have dementia syndromes (with
probable vascular aetiologies) in evolution.

Secondary Depressive Disorders


In addition to MRI investigation of elderly depressed patients, the clini-
cal and neuroradiological investigation of patients with secondary
depressions (e.g., post-stroke, Parkinson's disease [PD], Huntington's
disease [HD], MS, Alzheimer's disease [AD], cerebral tumour and au-
toimmune disorders) has significantly increased understanding of the
possible relationships between the neuroanatomical sites of pathology
and the occurrence of affective and/or cognitive symptoms (see Stark-
stein and Robinson 1993). Mayberg (1993) has elegantly described the
way in which results from studies utilising patients with depressive syn-
dromes secondary to subcortical diseases can be integrated to produce a
more coherent functional model of the pathophysiology of secondary de-
pressions. The specific relationships between depression in late-life and
neurological and neurodegenerative disorders characterised by mood,
movement and/or cognitive changes now require direct clinical, longi-
tudinal and imaging evaluation (see Tables 17.1-17.3).
The key questions are whether patients with melancholia share clinical
features, MRI changes, known risk factors and/or have a longitudinal
Magnetic Resonance Imaging 263
course similar to that of patients with known subcortical white matter
diseases (e.g., MS), basal ganglia disorders (e.g., PD and HD) and/or
other neurodegenerative disorders, particularly the multi-infarct demen-
tias and AD. Examination of such factors is the subject of intense current
investigation (see Tables 17.1-17.3). Within such a conceptual frame-
work, it is not simply proposed that melancholia (or late-life depression)
is just a variant of the more prevalent and well-described "cortical" and
"subcortical" dementias. Instead, melancholia is viewed as a disorder
which shares some common clinical features and, hence, is likely to share
some common pathophysiological processes (which may become the tar-
gets of specific treatment strategies) and, most importantly, some key
risk factors (which it may be possible to alter to prevent the disorder).
As discussed in Chapter 3, if clinical comparisons between such dis-
orders move away from simple descriptions of depressive sub-types, to
the detailed evaluation of (and interaction between) key depressive con-
structs (mood, cognitive impairment, psychomotor change and psychotic
features) then considerable new insights into the mechanisms underly-
ing such neurobehavioural constructs may be derived. Specifically, care
needs to be devoted to the evaluation of both structural and functional
neuroimaging studies in patients with comparable depressive disorders
occurring in the context of neurological disease. The patterns of dys-
function need to be closely related to the key behavioural constructs
described. Similarly, genetic and/or environmental risk factors specific
to these disorders need to be evaluated, with particular attention to the
likelihood that proposed risk factors may pre-dispose to some, but not
to other, clinically similar disorders. For example, although, Saunders
and Roses (1993) had claimed that the apolipoprotein E (ApoE) allele
E4 constitutes a specific risk factor to late-onset Alzheimer's disease, it
now appears that this same factor may also be a risk factor to late-onset
depression (Krishnan et al. 1994), and other neurodegenerative disorders
such as Creutzfeldt-Jakob disease (Amouyel et al. 1993). Such effects
may be mediated directly via CNS mechanisms (e.g., implicated role of
ApoE alleles in brain metabolism, with special relevance to the regula-
tion of glial cells in response to neuronal damage and deafferentation),
or indirectly, via increased risk to cerebrovascular disease.
In a series of studies, Robinson and colleagues (1984, 1988a, 1988b)
presented evidence that suggested that left-sided and frontal cerebrovas-
cular lesions were correlated with the onset of post-stroke depressive
disorders. Importantly, Starkstein and Robinson (1993) also demon-
strated the salience of left basal ganglia lesions for depression, and noted
264 /. Hickie, C. Hickie, E. Scott and K. Wilhelm
that strokes directly involving the caudate nucleus were strongly asso-
ciated with depressive disorders. Subsequent PET studies (Mayberg
1993) have confirmed that such lesions result in distant cortical effects.
Notably, selective caudate lesions (resulting in non-motor strokes) were
associated with focal changes in prefrontal cortex metabolism. Affec-
tive syndromes in general were associated with bilateral hypometabolism
throughout limbic regions (orbital-inferior frontal cortex, anterior tem-
poral cortex and cingulate cortex) with "the most pronounced changes
occur[ing] in the temporal lobes."
Patients with hypertension are at increased risk of developing sub-
cortical white matter changes, presumably as a consequence of chronic
hypoperfusion. Other vascular syndromes with particular relevance to
the CNS, such as the recently described anti-phospholipid antibody syn-
drome (APLS), require close evaluation in depressive patients. Corre-
spondingly, careful examination of the psychopathology of APLS pa-
tients with recognised neuropsychiatric sequelae is also warranted. Maes
et al. (1991) reported higher anticardiolipin autoantibodies (one type
of antiphospholipid autoantibody) in depressed patients compared with
control subjects. Extending the range of antiphospholipid antibodies
assessed in a subsequent study, Maes et al. (1993) concluded that "de-
pressives, and, in particular, melancholies, were significantly discrimi-
nated from normal controls by AC A (anticardiolipin), APTA (antipartial
thromboplastin) and APSA (antiphosphatidylserine) as discriminating
variables." In an MDU sample, we have noted a higher than expected
prevalence of IgG and/or IgM APL antibodies (25% of the sample, com-
pared with reports of 1% in normal young individuals and of 10-15% of
elderly community populations [Scott et al., unpublished data]). Fur-
ther, other autoimmune disorders which were thought to affect only pe-
ripheral structures, such as Sjogren^s syndrome, have now been demon-
strated as showing increased rates of subcortical hyperintensities on MRI
scanning, again suggesting hypoperfusion and neuronal loss.
Patients with diseases of the basal ganglia, notably HD and PD, have
been extensively evaluated neuropsychiatrically. In addition to the char-
acteristic movement disorders (with that seen in PD being phenotypi-
cally close to that observed in melancholia [e.g., Fleminger 1992; Sachdev
and Aniss 1994]), both these disorders are associated with affective syn-
dromes and cognitive impairment (Starkstein and Mayberg 1993; Peyser
and Folstein 1993). Cerebral tumours and closed head injury, particu-
larly where tissue damage involves frontal and temporal lobe structures,
Magnetic Resonance Imaging 265
have been consistently linked with depressive disorders (Lishman 1980;
Fedoroffet al. 1992).
Patients with MS may demonstrate extensive involvement of the sub-
cortical white matter, with total lesion area and corpus callosum atrophy
being correlated with the severity of global cognitive dysfunction, and
specific functions including slowed reaction time, impaired memory, con-
ceptual reasoning and inter-hemispheric information transfer (see Rao,
Leo and St. Aubin-Faubert 1989; Junque et al. 1990). Such correla-
tions are analogous to our findings in depressed patients (see Chapter 9)
and support the notion that hyperintense lesions in the white matter
on MRI are of functional significance. Depression is common in MS, as
is a range of other affective syndromes including bipolar disorder. As
with other secondary affective syndromes (e.g., multi-infarct dementia,
AIDS-dementia complex), however, the differentiation of true manic syn-
dromes from chronic or episodic behavioural disinhibition in MS patients
is contentious. In MS patients, affective syndromes appear to be most
marked in patients with MRI evidence of temporal lobe changes (Honer
et al. 1987; Ron and Logsdail 1989). The particularly high prevalence of
fatigue as a major disabling symptom in MS patients suggests also that
white matter lesions may be of particular relevance to our understanding
of the neurobiology of psychomotor disturbance.
Starkstein and Robinson (1993) drew an important conclusion with re-
gard to the secondary depressive disorders encountered in patients with
neurological disorders: although "several unifying ideas can be found ...
depression is not a unitary phenomenon. ... [T]here likely are multi-
ple types of depression within each neurologic disorder." Those authors
warn against over-interpretation of clinical similarities, but recognise
the way in which such insights might productively guide future depres-
sion research. For the secondary depressive disorders, the site of lesions
appears to be pertinent, with significance being attributed to anterior
(frontal and temporal) rather than to posterior lesions, to left rather
than to right-sided lesions (although right-sided lesions may be associ-
ated with mania) and for subcortical rather than cortical lesions.

Conclusion
MRI studies of patients with severe depressive disorders (typically occur-
ring in late life) and patients with depressions secondary to other neuro-
logical disorders have highlighted the prevalence, clinical relevance and
possible aetiological significance of abnormalities in the subcortical white
266 /. Hickie, C. Hickie, E. Scott and K. Wilhelm
matter and basal ganglia. For melancholia research, such studies support
the notion that depressions characterised by movement disorders (i.e.,
psychomotor change) and cognitive impairment may be underpinned by
structural changes in key fronto-striatal circuits (see Chapter 3). Our
ongoing research priorities with MRI now include:
• Correlational studies examining the relationships between the ex-
tent and location of specific MRI changes (e.g., reduced caudate and
putamen volumes, extent of limbic hyperintensities) and the style
and severity of key behavioural constructs (e.g., non-interactiveness,
psychomotor retardation and agitation, psychotic features, subcor-
tical and cortical neuropsychological impairments)
• Longitudinal studies, involving repeat clinical, MRI and neuropsy-
chological examination, to determine the relationships between melan-
cholia (particularly when its onset is in late life) and other relevant
neurological and neurodegenerative disorders.
18
Functional Neuroimaging in Affective
Disorders
MARIE-PAULE AUSTIN
PHILIP MITCHELL

Introduction
Melancholic depression is by nature an episodic disorder, with imputed
transient and/or fluctuating abnormalities in neurotransmission (Ferrier
and Perry 1992). Given the hypothesis (Austin and Mitchell 1995; Krish-
nan 1993b; and Chapter 15) that dysfunction in frontal-subcortical neu-
ral pathways contributes to the pathogenesis of melancholia, the use of
functional imaging, which allows us to capture the physiological changes
present at the time of scanning, is ideal for the study of such a recur-
rent clinical condition. Neuroimaging strategies can be divided into
those providing information on structure (e.g., CT and magnetic reso-
nance imaging [MRI]) or on function (e.g., positron emission tomogra-
phy [PET] and single photon emission computed tomography [SPECT]).
The introduction of functional imaging techniques such as PET and
SPECT to quantify changes in regional cerebral blood flow (rCBF) and
metabolism thus holds the potential to elucidate the neuroanatomical
substrate specific to melancholic (vs. non-melancholic) depression and
may even eventually aid in the diagnostic process.

PET and SPECT Techniques. PET allows assessment of regional


cerebral metabolism, and therefore regional cerebral function. In con-
trast to PET, SPECT measures rCBF, relying on the premise that un-
der normal conditions, rCBF is tightly yoked to regional metabolism.
Whereas SPECT measures the emission of single gamma rays or pho-
tons from the radiotracer, PET measures the simultaneous emission of
two gamma rays (produced by the reaction of an electron and a positron)
and thus enables "absolute" quantitation of gamma ray counts to take

267
268 M.-P. Austin and P. Mitchell
place. The need for a cyclotron to produce the positrons for PET does,
however, generate significantly higher costs than for SPECT.
The radionuclides most commonly used for the measurement of sin-
gle photon emission are 133Xe and "Tc-HMPAO, both of which allow
measurement of rCBF. The radionuclides used with PET are 18F-fluoro-
2-deoxyglucose (which measures metabolic rate) and 15O-labelled water
(which measures rCBF). When SPECT scanning is performed with a
multi-detector camera, it gives an image resolution which is comparable
to that obtained with state of the art PET scanners (i.e., structures as
small as 6-8 mm can be accurately visualised). The images produced by
PET and SPECT are tomographic (three-dimensional) and allow visual-
isation and measurement of rCBF in both cortical and subcortical struc-
tures. This is in contrast to non-tomographic CBF techniques (such as
are used with the radiotracer 133Xe), where the images produced are two-
dimensional and allow only for the measurement of cortical blood flow.
Whilst absolute quantitative analyses cannot be performed with
SPECT (unlike PET), in practice most studies involving either tech-
nology use relative or semi-quantitative analyses. Here, a region of in-
terest (ROI) is assessed relative to a reference region where cerebral
blood flow is held to be reasonably constant - usually the cerebellum
or visual cortex. The added advantage of such semi-quantitative analy-
ses (which, in effect, give ratios rather than absolute values of cerebral
blood flow) is to minimise the impact of fluctuations in global cerebral
blood flow on rCBF. The major advantage of SPECT (over PET) re-
mains its substantially lower cost, allowing for greater scanning numbers
and, consequently, improved statistical power.
In contrast to other neurobiological techniques used to study the ae-
tiological basis of depression, the great advantage of functional imaging
is that it allows the study of in vivo brain-behaviour relationships.

The Search for Abnormal Patterns of Metabolism and CBF


in Depression
While our research focusses on the melancholic type of depression, there
are very few imaging studies limited to melancholia. Presumably re-
flecting the dimensional view of depression, most have included hetero-
geneous depressive disorders, clouding interpretation of possible melan-
cholia-specific imaging changes. Where sub-groups of those with likely
melancholia were studied, we focus on reported findings for these sub-
jects.
Functional Neuroimaging in Affective Disorders 269
Resting Cross-sectional Studies
Of the larger and better controlled 133Xe studies, Sackeim et al. (1990),
who examined older subjects, and Delvenne et al. (1990), who scanned
a younger group, both reported reduced global cortical flow at rest in
their depressed patients, while Delvenne et al. (1990) reported greater
left hemisphere reduction in their endogenous group. Silfverskiold and
Risberg (1989), on the other hand, found no such difference between
depressed patients and matched controls. The major limitation of such
studies has been the inability to quantitate subcortical flows.
A number of 18FDG PET studies (examining small samples of depres-
sives) have found relative hypometabolism (greater on the left than the
right) in the frontal lobes (Baxter et al. 1989; Hurwitz et al. 1990; Mar-
tinot et al. 1990) and caudate nuclei (Baxter et al. 1985; Buchsbaum
et al. 1986; Hagman et al. 1990) of depressed subjects, thus support-
ing the possibility of prefrontal cortex and caudate dysfunction in these
patients.
Studies specifically investigating patients with endogenous or melan-
cholic depression have demonstrated abnormal perfusion in a number of
paralimbic structures. In a "Tc-HMPAO SPECT study undertaken in
Edinburgh comparing subjects (average age 45) with endogenous and
non-endogenous depression (as determined by the Newcastle depression
scale), Austin et al. (1992a) reported a positive correlation between New-
castle scores and frontal and cingulate rCBF in the endogenous group,
compared to the non-endogenous group in whom rCBF was decreased
both cortically and subcortically. A recent "Tc-HMPAO study by May-
berg et al. (1994), examining 13 subjects under age 55 with severe unipo-
lar depression, reported significant reduction in frontal cortex, anterior
temporal cortex, anterior cingulate cortex and caudate nuclei compared
to a matched control group. Interestingly, psychiatric symptom rating
scale scores did not correlate with rCBF, whilst scores on a scale assess-
ing psychomotor slowing - the Unified Parkinson's Disease Rating Scale
(Fahn and Elton 1987) - correlated negatively with frontal and cingulate
perfusion.
An 15O PET study of 33 "severely depressed" elderly patients (Bench
et al. 1992) - many of whom, presumably, were melancholic - showed
reduced rCBF in the left dorsolateral prefrontal cortex and anterior cin-
gulate in the group as a whole, whilst the more cognitively impaired
sub-group had greater rCBF reduction in the medial frontal cortex. A
subsequent study by the same research group (Bench et al. 1993b) ex-
270 M.-P. Austin and P. Mitchell
amined for correlations between patients' symptom ratings (which had
been factor analysed) and patterns of rCBF. The psychomotor retar-
dation factor correlated negatively with rCBF in the dorsolateral pre-
frontal cortex (DLPFC), a finding consistent with the Mayberg et al.
1994 study discussed above, whilst that for cognitive performance corre-
lated positively with rCBF in the left medial prefrontal cortex. A recent
review of 100 depressed subjects scanned with "Tc-HMPAO SPECT by
the Edinburgh group (Goodwin 1994, personal communication) demon-
strated small but significant reductions in left frontal and bilateral cau-
date rCBF. This group also reported significant positive correlations
between left frontal flows and endogeneity (as assessed by the Newcastle
scale) in the under 40 group, while the reverse held true for the over 60
group.

Follow-up Studies
The major problem with cross-sectional studies is the potential for con-
founding state and trait abnormalities. Rescanning at recovery is thus
important to circumvent this problem, and if undertaken in conjunction
with MRI, allows for better identification of the relationship between
changes in CBF/metabolism and concurrent structural abnormalities.
Martinot et al. (1990), in an 18FDG PET study of 10 severely de-
pressed patients matched for medication status before and after recov-
ery, reported disappearance of left-right asymmetry, but persistence
of hypofrontality, upon recovery. Baxter et al. (1989), on the other
hand, showed normalisation of relative left DLPFC hypometabolism af-
ter treatment with antidepressants. Bench et al. (1993a), rescanning 25
of their original 33 depressed patients, found that remission was associ-
ated with significant normalisation in left DLPFC rCBF but not in the
cingulate and angular gyri. In contrast, Goodwin et al. (1993), using
"Tc-HMPAO with SPECT, rescanned 16 previously depressed subjects
(matched for medication status pre- and post-recovery) and found that,
upon recovery, frontal cortical perfusion deficits persisted, whilst there
was a return to normal perfusion in limbic and subcortical structures
(basal ganglia, anterior and posterior cingulate and thalamus). They
concluded that the persistent cortical reduction in rCBF either repre-
sented a trait abnormality or was simply related to concurrent volumet-
ric reduction (Coffey et al. 1993) and/or white matter changes (Dolan
et al. 1990; Coffey et al. 1993) in the frontal cortex of depressed patients.
The pattern of rCBF state changes (in dopaminergic-rich regions) might
Functional Neuroimaging in Affective Disorders 271
implicate the involvement of dopaminergic pathways in the pathogene-
sis of depression. Drevets et al. (1992), using 15O PET, compared 13
unmedicated depressed young unipolars with 10 matched subjects in re-
mission, and found a significant increase in rCBF in left ventrolateral
and medial prefrontal cortex in the depressed subjects and a trend for
left amygdala rCBF increase in both groups. They concluded that in-
creased activity in the left prefrontal cortex was a state abnormality
and that the abnormality in amygdala activity might represent a trait
abnormality.
In conclusion, although both state and trait changes have been iden-
tified in these follow-up studies, there is a lack of consensus as to the
exact pattern of these rCBF changes in melancholia. Better definition
of clinical study samples, concurrent co-registration with MRI and the
use of activation studies may all be useful in more clearly identifying the
nature of such abnormalities.

Activation Studies
The potential problems associated with studies performed on subjects
in the resting state (such as those reviewed above), and the advantages
of activation studies, have been well described by Gur, Erwin and Gur
(1992). A particular problem in resting studies is the lack of ability to
control adequately the resting state, leading to a potential lack of homo-
geneity in patient behaviour at the time of scanning, and thus increased
variance in measures of rCBF. The use of activation paradigms, whereby
all subjects are demonstrably performing the same task, is one way of
circumventing this problem. Furthermore, when applied to psychopatho-
logic conditions, the use of neurobehavioural and neuropharmacological
activation "probes" may be able to link regions of abnormal physiologic
activation with specific behavioural deficits. Given that resting changes
in rCBF in patients with functional psychiatric disorders are small (al-
beit statistically significant), the activation method of scanning holds
the potential for enhanced detection of abnormal patterns of rCBF.
To date, only one well-controlled activation study has been performed
in depressed patients. In a 133Xe study, Berman, Pickar and Weinberger
(1993), using the Wisconsin Card Sorting Test (WCST) as a neurocog-
nitive probe to highlight regional CBF in the prefrontal cortex (PFC),
scanned 10 severely depressed unmedicated subjects, 10 with chronic
schizophrenia, and 10 matched controls. In contrast to the group with
schizophrenia, who were unable to activate the PFC upon performance
272 M.-P. Austin and P. Mitchell

of the WCST, they found no significant difference between depressed


subjects and matched controls in terms of the ability of the WCST to
augment prefrontal CBF. Unfortunately, the 133Xe technique does not
allow for study of striatal CBF, abnormalities of which have been previ-
ously noted in depressed subjects at rest. This methodological limitation
may well have obscured differences between depressed and control sub-
jects at the subcortical level.

Methodological Problems
The conflicting findings reported in the resting studies reviewed above
may be related to a number of potential methodological difficulties inher-
ent in this area of research. Of great importance is the lack of diagnostic
homogeneity (a point which will be taken up in the discussion), which
can obscure otherwise significant changes in rCBF. In addition, age,
medication, anxiety at the time of scanning, and concurrent structural
cerebral abnormality can all affect patterns of rCBF.
Advancing age has been associated with a decline in frontal and sub-
cortical CBF, whilst cortical atrophy is most closely associated with
reduced global CBF (Waldemar et al. 1991). The use of adequately age-
matched control groups is thus crucial, and co-registration with MRI
(Kosugi et al. 1993) to allow the assessment of interaction between struc-
tural and functional abnormalities in depression is highly advisable. The
potential contribution of structural changes to functional abnormalities
seen at baseline can be assessed by rescanning patients in the recovered
state. If baseline rCBF changes have normalised upon recovery, this
suggests that changes in the pattern of rCBF during the depressed state
reflect abnormalities in regional brain activity related specifically to the
state of depression rather than being secondary to structural changes in
these and other functionally related regions.
The effect of psychotropic medication, in particular antidepressants,
on rCBF has been poorly documented, and must always be considered
as a potential confounding variable in such studies. Furthermore, the
effects of other drugs need to be controlled for as much as possible,
as it has now been reported that long-acting benzodiazepines, caffeine
and chronic smoking are all associated with global reductions in CBF
(Mathew, Wilson and Daniel 1985). Anxiety has been reported to have
a significant effect on global CBF by some researchers (Gur et al. 1987)
but not by others (Giordani et al. 1990).
In the case of activation studies, one of the major potential con-
Functional Neuroimaging in Affective Disorders 273
founders with cognitive paradigms is the individual's ability to perform
the required task during scanning (Sergent 1994). In order to control for
performance factors, it is thus crucial to use activation tasks that can be
performed by even the least able subject. Central to activation studies is
the subtraction method, whereby (in the case of cognitive activation) the
rCBF associated with a control task is subtracted from that associated
with the experimental task. In this way the non-relevant (e.g., motor,
sensory) aspects of the cognitive task can be cancelled out, allowing the
impact of the purely cognitive task on rCBF to be assessed. To do this,
there must be very close matching between control and experimental
tasks (Gur et al. 1992).

The Diagnostic Utility of Functional Imaging:


Is It Currently Justified?
The above studies report relatively subtle changes in rCBF when aver-
aged means are obtained and compared across reasonably large groups of
patients. The power of the matched control group study, where between-
group differences are examined and quantified, allows such changes to
be detected. It is, however, much harder to demonstrate these subtle
abnormalities in the individual case. As a result, the application of func-
tional imaging techniques to the diagnosis of neuropsychiatric disorders
remains problematic.
There have been two published assessments of functional scanning
in the clinical setting (George et al. 1991; Parker and Austin 1995),
with both reports suggesting that SPECT (with the use of visual scan
analysis) may have an important role in the assessment of individual
patients in a number of different clinical settings. It is likely, however,
that the clinical utility of functional imaging in the neuropsychiatric
setting will be limited.
One important potential diagnostic use of functional neuroimaging is
to distinguish between those with early Alzheimer's dementia (AD) and
those with a melancholic depression who present with cognitive distur-
bances suggestive of a dementia ("pseudo-dementia"). To date, only
a small number of studies have directly compared differences in rCBF
between matched depressed and AD patients. Curran et al. (1993), us-
ing "Tc-HMPAO with a high-resolution SPECT scanner, found global
hypoperfusion in both their depressed and demented subjects, with the
degree of reduction in global CBF being less in depressed subjects com-
pared to their demented counterparts. They also reported a significant
274 M.-P. Austin and P. Mitchell
reduction in temporo-frontal, anterior cingulate and subcortical regions
in their depressed male subjects compared to a group of matched con-
trols. Perfusion in the posterior parietal cortex in these depressed sub-
jects was, unlike their AD counterparts, predominantly normal. Sack-
eim et al. (1993), in their 133Xe study, found that matched elderly de-
pressed and demented subjects had equivalent reductions in global CBF,
with different patterns of regional cortical hypoperfusion differentiating
the two patient groups. Grunwald et al. (1993), in a qualitative multi-
detector SPECT study, found bilateral reduction in parieto-temporal
perfusion in 35% of AD subjects without such changes in any of the
matched depressed elderly group. They also concluded that this pattern
of under-perfusion, when present, could be used to exclude depression.
Upadhyaya et al. (1990) found reductions in global CBF in the two
groups similar to those reported by Curran et al. (1993). The limited
resolution of their SPECT scanner and the use of qualitative scan anal-
ysis did not allow for evaluation of changes in regional CBF; therefore,
differentiation between the two patient groups was not feasible. Finally,
Kumar et al. (1993), in a small preliminary 18FDG PET study compar-
ing AD, depressed and matched control subjects, reported widespread
metabolic reduction in most cortical and subcortical regions in their de-
pressed group, comparable to those seen in the demented (AD) group
and absent in the control group.

The confidence with which brain SPECT scans can be visually (qual-
itatively) interpreted for the purpose of assisting diagnostic interpreta-
tion in individual cases is directly related to the degree of reduction in
CBF in specific regions. Thus, in Alzheimer's dementia, where there
are relatively large reductions in cerebral blood flow, visual analysis
of the scans alone will be reasonably accurate in identifying perfusion
deficits over and above any inter-subject variability that may be present.
When one analyses scans from patients with primary psychiatric disor-
ders such as depression, the changes in rCBF would be expected to be
much smaller. In such instances, it may be much more difficult to iden-
tify visually reductions in rCBF. Therefore, the use of semi-quantitative
(relative) analysis and the availability of pooled control data with which
to compare patient data becomes important.

The studies reviewed nevertheless suggest that SPECT may be a use-


ful diagnostic adjunct in differentiating between the dementias and de-
pressive pseudo-dementia.
Functional Neuroimaging in Affective Disorders 275
Diagnostic Specificity in Functional Imaging Studies
of Depression
Are there any rCBF findings specific to depression? A number of studies
have addressed this issue. For example, both Buchsbaum et al. (1986)
and Cohen et al. (1989) have reported a reduction in glucose metabolism
in the DLPFC of schizophrenic and depressed subjects, and concluded
that both disorders may be mediated by the same brain mechanism.
Dolan et al. (1993), in their 15O PET study, reported that patients with
poverty of speech (as an indicator of psychomotor poverty, which is
common to both schizophrenia and melancholia) had significantly lower
rCBF in the left DLPFC irrespective of diagnosis. They concluded that
clusters of clinical features may be more closely related to patterns of
rCBF than diagnostic labels. While this may appear to be true if one
is looking at the prefrontal cortex as a whole, it is quite possible that
dysfunction in different areas/systems of the PFC produces the different
symptom manifestations in these two disorders. Increasing awareness of
the presence and complexity of frontal-subcortical networks (Alexander,
DeLong and Strick 1986; Alexander, Crutcher and DeLong 1990) and
their relevance to a number of different functional psychiatric disorders
indicates the possibility that studies which do not examine more cir-
cumscribed regions of the PFC are unlikely to allow identification of
discrete abnormalities in patterns of CBF/metabolism in different pa-
tient populations. Fortunately, the degree of spatial resolution available
with the current generation of PET and SPECT scanners now allows for
investigation of more discrete regions of the PFC.

Conclusion
The studies reviewed above strongly implicate the prefrontal cortex and
basal ganglia in the pathophysiology of melancholia, supporting involve-
ment of putative frontal-subcortical networks in the pathogenesis of
melancholia (as discussed in Chapter 15). Further studies in better-
defined groups of melancholic patients should lead to further refinements
in our understanding of these pathophysiological mechanisms, and as a
result, improved pharmacological treatments for melancholia.

Future Directions
One of the most important tasks in improving the power of imaging
studies is to maximize the homogeneity of the study group, thereby
276 M.-P. Austin and P. Mitchell
highlighting any true abnormalities in the pattern of rCBF. The use of
a sign-based instrument such as the CORE measure should be a most
useful tool - not only in obtaining better-defined samples of melan-
cholic subjects, but also in differentiating between sub-samples on the
basis of a behavioural feature (psychomotor disturbance) which appears
to have specificity to melancholia. Additionally, the use of rescanning
upon recovery becomes crucial if one is to distinguish trait from state
abnormalities.
Appropriate motor, neuropsychological and neuropharmacological ac-
tivation paradigms, aimed at targetting the psychomotor abnormalities
so central to melancholia, are also likely to be valuable. More sophisti-
cated statistical analyses such as the use of pixel-by-pixel analysis (Fris-
ton et al. 1991) and correlational analyses will allow a more sensitive
assessment of the subtle patterns of physiological dysfunction present in
melancholia. Co-registration of MRI and PET scans (Kosugi et al. 1993)
should allow not only improved anatomical localisation of regions of in-
terest in functional imaging studies, but also a greater understanding
of the relationship between structural and functional deficits in melan-
cholia. Finally, the isolation of new dopamine receptors (Kapur and
Mann 1992) - in particular dopamine D3, D4 and D5 receptors, which
predominate in limbic areas - may further clarify the potentially impor-
tant modulatory role of dopaminergic pathways in the pathogenesis of
melancholia. Further imaging studies using dopamine receptor radioli-
gands should greatly aid in this endeavour.
19
Summary and Conclusions
GORDON PARKER
DUSAN HADZI-PAVLOVIC

As we have presented our arguments and empirical analyses in some


detail in the preceding chapters, this will be a succinct overview of the
issues and conclusions.
We have noted (Chapters 1-3) some major limitations to recent and
current formal classifications of the depressive disorders, particularly
limitations in a number of assumptions, which, by forcing homogeneity
across heterogeneous conditions, almost certainly provide invalid defi-
nitions of melancholia. As a corollary, such limitations have restricted
identification and quantification of aetiological factors and of treatment
specificity to separate depressive disorders to an extent that prejudices
treatment. Such limitations restrict the application of the increasingly
sophisticated tools and technologies now available to the biological psy-
chiatrist. Strauss (1994) has similarly questioned whether biological
psychiatry is building on an adequate base. He stated that it now seems
that "with impressive and increasing sophistication in the neurosciences,
we are leaving patient description far behind." He argued that "current
categories" may have been adequate to begin to understand disorders,
but that "such categories may reflect simplistic, static notions that relate
only very approximately to the core processes involved in psychopathol-
ogy." Further, he suggested that if we refuse to attend to the phenomena
that reflect core processes, "Biology is going to be severely handicapped
in finding biological connections to psychopathologic mechanisms." Such
concerns underpinned our research objectives and have therefore been
detailed extensively.
Viewing melancholia as a psychiatric disorder highly amenable to cir-
cumscription, we have sought to identify its denning clinical features.
We have reviewed the utility of psychomotor disturbance (PMD) to
depressive disorders. Rather than view it as merely one among many

277
278 G. Parker and D. Hadzi-Pavlovic
clinical features defining depressive severity, we have considered its util-
ity as a feature specific to the "melancholic" type of depression. We have
argued on the basis of empirical analyses (principally the application of
latent class analyses [LCAs]) that demonstration of its specificity and its
utility requires rating observed PMD rather than relying on symptom
reports.
We have described (Chapter 6) the progressive development of a clinic-
ian-judged measure of PMD (the CORE), and, in applied studies (Chap-
ters 6-13), noted several ways by which CORE scores can be used in
categorical and dimensional analyses. We provide specific details on pro-
cedural issues and rating (Chapter 14 and Appendix) of the final 18-item
CORE measure, and note (Chapter 6) its interesting "trunk and branch"
structure - with a truncal non-interactiveness (or cognitive processing)
component as well as arborising retardation and agitation components.
We have demonstrated (Chapter 6) a number of limitations to his-
torically suggested "endogeneity" symptoms as markers of melancholia,
with most appearing to more define depression and depression sever-
ity rather than any depressive sub-type. By contrast, we argue that
CORE-defined PMD does possess such specificity - being absent in non-
melancholic depressive disorders and present and noteworthy (albeit
varying in severity) in melancholic depression. We have sought to revive
interest in PMD, a clinical variable long respected by many clinicians.
In the polemic Culture of Complaint, Robert Hughes observed that "art
advances by injecting doses of unacceptability into its own discourse."
By contrast, we suggest that medical research is most likely to advance
when any quintessential clinical acceptables are respected.
We suggest that the potential utility of CORE-defined PMD as a marker
of melancholia appears striking. We constructed an excessively parsimo-
nious hypothesis (i.e., that PMD is both a necessary and sufficient fea-
ture of melancholia). Thus, we examined whether all those with "melan-
cholia" reached criterion status for PMD and that no other endogeneity
feature was necessary (presumably being subsumed by PMD). While the
hypothesis did not achieve absolute support - in that a few symptoms
were suggested as providing an additional mantle to the identification
of melancholia - the proposition received stronger support than antic-
ipated (especially when confounding issues such as measurement error
and severity influences must be conceded).
We suggest that not only does the CORE system appear to be capable
of defining neurobiologically discrete groups but, to the extent that it
actually defines "melancholia," then it provides a long-sought marker of
Summary and Conclusions 279
a neuropathological process underpinning melancholia. This is of some
importance when research findings have generated a number of impor-
tant models (especially those by Carroll 1991; Cummings 1993; Krishnan
1993b; Mayberg 1993; and Austin and Mitchell [1995; and Chapter 15])
implicating both functional perturbations and structural disturbances
to fronto-subcortical neural networks. As detailed in Chapter 3, the
CORE measure identifies a key behavioural construct to melancholia and
provides an investigatory strategy to proceed with important neurobio-
logical research - whether involving a single modality or, as detailed by
us (Chapters 3 and 15), multiple investigatory strategies and a "conver-
gence of evidence" approach.
We have considered the properties of the CORE measure extensively.
Two reliability studies (Chapter 6) suggest that, at least in the hands
of trained clinicians (not necessarily psychiatrists), high inter-rater re-
liability can be achieved - even when ratings are based on videotaped
interviews. In terms of validity, we pursued the CORE'S capacity to de-
fine melancholia, and therefore examined its capacity to differentiate
some of the givens of that condition. Those assigned (Chapter 6) by the
CORE I measure as having putative melancholia were less likely to re-
port early or current deprivational experiences, and less likely to report
recent or chronic depressogenic stressors. In addition, we demonstrated
that those CORE scores predicted response to tricyclic medication (Chap-
ter 10). The CORE II validity studies were necessarily more extensive.
We established differentiation using the DST as a biological index of
melancholia (Chapters 6 and 8). We also established that CORE-assigned
melancholic depressives were older, older at first episode of depression,
more severely depressed, more likely to report clear remissions to pre-
vious episodes and more likely to have psychotic features (Chapter 6).
In addition, we demonstrated strong links with neurocognitive test re-
sults (Chapters 6 and 9), with bimodal distributions arguing against
the unitary model of depression. Refined analyses indicated that our
conclusions were not confounded by subjects with psychotic depression,
or that CORE scores were merely defining a more severe expression of
depression. In terms of outcome studies, we report an important ECT
study (Chapter 10) strongly supporting the predictive validity of the
measure, but acknowledge that more extensive and sophisticated stud-
ies are required (particularly of antidepressant drug response) to assess
one of the givens of melancholia (i.e., that it is selectively responsive to
physical treatments).
We developed an empirically driven clinical algorithm for defining
280 G. Parker and D. Hadzi-Pavlovic

melancholia (Chapter 13) building in a small set of refined clinical fea-


tures, and demonstrated its comparative utility (against other melan-
cholia indices) and its likely validity.
We have addressed two additional phenomenological objectives. One
involved the clear demonstration (Chapter 11) that those with unipolar
and bipolar melancholia have similar clinical features during the melan-
cholic phase. Additionally, we have pursued the clinical definition and
status of psychotic depression (Chapter 12). The latter studies provide
support for the view that psychotic depression is a sub-class of melan-
cholia (see below) and that PMD is a key feature of that disorder. We
provide strong evidence to suggest that psychotic depression may exist
in some patients for whom delusions and hallucination have not actually
been elicited, and explore the implications of that important finding.
We address (but do not completely resolve) the unitary/binary ar-
gument. We suggest (Chapters 1-4) a number of reasons as to why
the binarian argument has failed to convince the nosologists (e.g., inap-
propriate or limited statistical approaches, reliance on symptom reports,
intrinsically differentiating clinical features being "swamped" by severity
measures). Our analyses (both of CORE II scores in Chapter 6 and the
latent class analyses reported in Table 12.3) suggest more a three-class
solution than a binary one. That is, when "noise" (generally reflected by
severity measures) is removed from any intrinsically differentiating set
of clinical variables, we find evidence of three sub-populations of depres-
sive disorders. These may reflect quite separate classes, or a hierarchy.
The latter model can be readily illustrated. We presume the lowest-
order hierarchical class to subsume an heterogeneous mix of depressive
conditions, but certainly not melancholic or psychotic depression, and
with features principally defining a mood disorder. At the next level is
melancholia, with its mood component also defined by a range of symp-
toms but with the presence of CORE-defined PMD being the additional
specific feature defining class membership. At the next level is psych-
otic depression, with all the features of the underlying mood disorder
and of melancholia (but with the PMD, along with several endogeneity
symptoms, being more severe) and with the additional specificity being
provided by morbid over-valued ideas, delusions and hallucinations. The
hierarchical model is appealing, as it concedes a cascading set of neu-
robiological domains being progressively involved, and may be tested in
future research as outlined in Chapter 3.
In a series of chapters (Chapters 3 and 15-17) we review recent re-
search identifying structural and functional neurobiological processes
Summary and Conclusions 281
implicated in the pathogenesis of melancholia, and consider in some
depth (Chapter 16) the possible relevance of the ageing brain as a real
neuropathological factor as opposed to a confounding influence. Sepa-
rate chapters examine structural brain abnormalities (e.g., volumetric
reduction in basal ganglia; white matter hyperintensities revealed in re-
cent MRI studies) and functional disturbances (e.g., abnormal patterns
of metabolism and regional cerebral blood flow revealed in PET and
SPECT studies), which, when linked with illness variables (e.g., fam-
ily history; age of onset) and other determinants, will allow a range of
diathesis-stress models to be pursued in the next decade.
We earlier noted Strauss' concerns about the inadequate clinical de-
scriptive base for biological psychiatric research. He concluded that "the
sophistication of biological psychiatry has far outstripped the sophisti-
cation of our research observations of patients," and that the "Decade
of the Brain for psychiatry may only be as successful as the adequacy
of our clinical assessment will allow." We suggest that application of
the CORE measure (in conjunction with investigation of several other
domains suggested by our refinement studies) may be an advance on
current diagnostic criteria sets in clarifying candidate neurobiological
processes involved in melancholia.
Appendix
The CORE measure: procedural
recommendations and rating guidelines

General Guidelines
This form should be completed only for patients suffering from
a depressive disorder. It aims to record important behavioural
aspects of such patients. Rate, therefore, on the basis of your
observations, rather than relying on the patient's description.
Ignore attributes of the patient's personality.
Ratings
0 always indicates the absence or triviality of a feature

indicate that the feature is distinct and pathological


3J
Procedural Recommendations
(1) You should undertake your normal clinical intake interview, taking
particular note of the patient's interactions. We suggest that a
minimum of 20 minutes is usually required before making these
behavioural ratings.
(2) For retardation items, rate conservatively, reflecting an overview
of the whole interview after the initial "settling in" phase. Thus, a
depressed patient with retardation may still show variable severity
of features. For example, at varying times, speech may be slightly,
moderately or severely slowed. In such instances, you should rate
the average level across the whole interview and, if unsure which

282
The CORE measure 283
of two scoring options to select, rate conservatively by choosing the
less severe option.
(3) For agitation items, the rating rules differ. Some depressed pa-
tients may present in a retarded way for much of the interview and
show agitation only for brief epochs, if at all, perhaps by briefly
getting up from a chair, pacing around, wringing their hands or im-
portuning, e.g., "What is to become of me?" While such features
may be expressed only briefly, they should be rated as categorically
present for relevant items, and with severity ratings reflecting the
average level of the expressed behaviour over one or more epochs.
False positive levels of retardation or agitation may occur in those
with significant anxiety or with a limited emotional range. Anxi-
ety, for instance, may result in a rather "frozen" facial or body
appearance, or be expressed in increased motor activity (such as
hand movements and restlessness). To clarify the base level, the
interviewer should adopt an interactive and reassuring style and so
help the patient relax, and make ratings only after a settling period
of 20 minutes or so. The base level of retardation, with its distinct
quality, is generally maintained and influenced minimally, if at all,
by the interviewer's interactive attempts. By contrast, any pseudo-
retardation effects due to anxiety should be less or no longer evident
as the interview progresses.
Agitation has some characteristic features. As the outward mo-
tor expression of mental perturbation, it is generally associated with
mental activity reflected behaviour ally by bewilderment, perplexity
and/or puzzlement, and by accounts of mental preoccupation with
extremely distressing issues which, to the observer, may often be
relatively trivial. Agitated patients generally appear impervious to
reassurance and even to interaction with the interviewer, with the
base level intensity of expression persisting or influenced only mini-
mally or briefly by distraction, interruption or the interviewer's reas-
surance and interactive warmth. The anxious or stressed individual
will, by contrast, tend to relax and settle any "pseudo-agitated"
movements as the interview progresses.
284 Appendix

Rating Guidelines
1 Non-interactiveness
For what proportion of the interview does the patient not respond to the
social cues or fail to interact with or "stay with" the interviewer, i.e., give the
impression that the interviewer has "not been registered"? A non-interactive
patient does not necessarily show an impairment of concentration.
0 —Consistently interactive

1 —Not interactive some of the time


2 —Not interactive much of the time
3 —Not interactive almost all of the time

2 Facial immobility
The rater should assess the lack of moment-to-moment fluctuation of facial
expression. The depth of expression is important such that shallow, fleeting
changes or mere social reactions should not be given great weight. Disregard
natural blinking in your assessment.
0 —Mobility within normal limits

1 —Somewhat restricted facial mobility


2 —Moderately restricted facial mobility
3 —Fixed and immobile face

3 Postural slumping
Judge the presence and severity of postural slumping (i.e., head bowed, shoul-
ders rolled forward) relevant to the patient's age and physical status, while
the patient is sitting, standing and walking.
0 —No slumping

1 —Slightly slumped posture


2 —Moderately slumped posture
3 —Markedly slumped posture
The CORE measure 285

4 Non-reactivity
Assess any failure by the patient to show improvement in mood in response
to something pleasing, or to your attempts at cheering the patient up. If
there is no spontaneous reactivity, test for it formally (e.g., ask about an
interest or normally pleasurable event, compliment the patient about some
characteristic or achievement or use humour). All patients who have failed to
smile during the interview should be asked formally to smile. Spontaneous or
unforced smiles, and full smiles evoked by request, should generate a 0 rating
for non-reactivity. Superficial or forced smiles favour a positive rating.
0 —Appropriately reactive mood

1 —Slightly non-reactive mood


2 —Moderately non-reactive mood
3 —Severely non-reactive mood (neither worsening nor
improving)

5 Facial apprehension
Rate the extent to which the patient's face shows sustained morbid apprehen-
sion, perplexity, bewilderment, fearfulness or tortured concern. The appre-
hension is unable to be relieved substantially by the interviewer's attempts to
provide realistic comfort or reassurance. The item should not be rated unless
the apprehension is clearly pathological and persistent.
0 —No facial apprehension

1 —Slight facial apprehension


2 —Moderate facial apprehension
3 —Marked facial apprehension
286 Appendix

6 Delay in responding verbally


Judge the extent to which the patient shows delay or inhibition in replying to
questions. Allow for the patient's education, culture, age and language.
0 —No obvious delay in responding

1 —Slight delay in responding


2 —Moderate delay in responding
3 —Severe delay in responding

7 Length of verbal responses


Rate the extent to which the patient refrains from speaking or responding at
any length in reply to more open-ended questions. Allow for the patient's
education, culture, age and language.
0 —Responses of appropriate length

1 —Responses distinctly shortened


2 —Responses generally of a few words only
3 —Mute

8 Inattentiveness
Inattentiveness is, in effect, an impairment of concentration as judged by the
observer. Rate the extent to which the patient is inattentive to the interview
and to the interviewer. The patient may have full consciousness but be inat-
tentive. This differs from non-interactiveness (Item 1) in that the patient may
well appear to interact but be unable to sustain attention to the interview.
0 —Consistently attentive

1 —Inattentive for some of the time


2 —Inattentive for much of the time
3 —Inattentive almost all of the time
The CORE measure 287

9 Facial agitation
Judge the extent to which the patient's facial movements and fluctuations
in expression indicate pathological fearfulness, bewilderment, anguish, per-
plexity or mental torment. Agitation can be commonly expressed in sudden
outbursts of anguish or despair. At other times the patient's face may lack mo-
bility. Do not rate on the basis of the patient's complaints of anxiety. Do not
rate dyskinetic movements which are not associated typically with tormented
mental features. Do not rate tremulous movements or physical disorders which
may produce apparent tremors. In distinguishing from movements associated
with anxiety, refer to Point 3 in the general guidelines. A 3 rating requires
persistent and significant agitation and/or several epochs of severe agitation
superimposed on a facial expression of bewilderment, perplexity and/or retar-
dation.
0 —No facial agitation

1 —Slight facial agitation


2 —Moderate facial agitation
3 —Persistent and/or several epochs of marked facial
agitation

10 Body immobility (amount, not speed)


Judge the extent to which the patient moves limbs, hands and body, relevant
to the patient's age and physical status.
0 —Mobility within normal limits

1 —Slightly restricted mobility


2 —Moderately restricted mobility
3 —Virtually no movement (immobile)
288 Appendix

11 Motor agitation
Rate persistent, excessive or inappropriate motor activity as manifested by a
characteristic inability by the patient to sit or stay still, indicating thwarted or
misdirected energy. Typical movements include slow rubbing, pacing, writhing
or wandering actions. The movements may have an autistic quality. Do not
rate tremors, dyskinesia or mannerisms. Note Point 3 in the general guidelines.
A 2 rating could reflect persisting agitation of moderate severity or epochs of
quite severe agitation, while a 3 rating reflects persistent and severe agitation.
0 —No abnormality, or movements more typical of anxiety

1 —Slight motor agitation


2 —Persistent agitation of moderate severity or epochs of
moderate intensity
3 —Severe motor agitation, unable to sit still at all

12 Poverty of associations
Judge the vagueness of information given, and the extent to which topics and
themes lack explication or richness, again allowing for education and cultural
differences. Ignore the extent to which the patient talks freely or not. Rate
in response to both open-ended and specific questions.
0 —No abnormality

1 —Slight poverty of associations


2 —Moderate poverty of associations
3 —Severe poverty of associations
The CORE measure 289

13 Slowed movement (speed, not amount)


Rate any slowing in movement, relevant to the patient's age and physical
status. Observe the patient entering and leaving the room, and sitting and
standing, or test by asking the patient to walk the length of the room.
0 —Normal speed of movement

1 —Slightly slowed movements


2 —Moderately slowed movements
3 —Severely slowed movements

14 Verbal stereotypy
Rate the extent to which the patient's speech is morbidly repetitive or per-
severative (e.g., constantly seeking but generally inaccessible to reassurance,
often importuning or complaining without heed to time or place). The repeti-
tion of limited themes, often of a mundane or seemingly inappropriate nature,
is particularly characteristic. Typically, the effect of interruption on such ver-
bal output is minimal, as is the degree to which the patient's themes are
altered or diverted by the interview situation.
0 —No obvious verbal stereotypy

1 —Slight verbal stereotypy


2 —Moderate verbal stereotypy
3 —Severe verbal stereotypy
290 Appendix

15 Delay in motor activity


Judge the extent to which the patient shows delay or inhibition in initiating
movement (e.g., getting up to leave the room), relevant to the patient's age
and physical status.
0 —No obvious delay in motor activity

1 —Slight delay in motor activity


2 —Moderate delay in motor activity
3 —Severe delay in motor activity

16 Impaired spontaneity of talk


Judge impaired spontaneity of talk after allowing for any likely educational,
cultural or other relevant influences, and rate in response to both open-ended
and specific questions.
0 —Spontaneously raises issues

1 —Slight loss of spontaneity


2 —Moderate loss of spontaneity
3 —No spontaneous talk

17 Slowing of speech rate


Judge any slowing in speech rate, allowing for the patient's age and physical
status.
0 —Normal rate of speech

1 —Slightly slowed rate of speech


2 —Moderately slowed rate of speech
3 —Markedly slowed rate of speech
The CORE measure 291

18 Stereotyped movements
Rate the presence and severity of certain persistent, repetitive or purposeless
movements. Stereotypic movements are distinct, unusual, idiosyncratic and
often bizarre. Examples of stereotyped movements might be rubbing of hands
in a rather bizarre manner, picking at the skin, fiddling with clothes and skin
in association with an autistic and perturbed mien. To be rated as present,
the movements should be obvious, characteristic and little affected by external
distraction. We regard "stereotyped movements" as a higher-order variable of
"motor agitation" (Item 11). Thus, all depressed patients with stereotyped
movements will have "motor agitation," but the reverse is not necessarily true.
If Item 11 ("motor agitation") has returned a 0 rating, then a 0 rating must
be returned here.

0 —Normal non-stereotyped movements

1 —Slight degree of stereotyped movements (use for


those who show hand-wringing only)
2 —Moderate degree of stereotyped movements
3 —Severe degree of stereotyped movements

Circle whether the mental state examination was undertaken near to or at the
nadir (most severe period) of the episode.
Nadir: Yes No

Calculating scale and total item scores


Non-interactiveness: 1 + 4+ 7 + 8

Retardation: 2 + 3 + 6 + 10 + 13 + 15 + 17 =

Agitation: 5 + 9 + 11 + 14 + 18 =

Total CORE score: (Total of all items) =


References

Abas, M. A., B. J. Sahakian and R. Levy (1990). Neuropsychological deficits


and CT scan changes in elderly depressives. Psychological Medicine 20,
507-520.
Abraham, K. (1949 [orig. 1911]). Notes on the psycho-analytic investigation
and treatment of manic-depressive insanity and allied conditions. In
Selected Papers of Karl Abraham. London: Hogarth Press.
Abrams, R. and M. A. Taylor (1974). Unipolar and bipolar depressive
illness. Archives of General Psychiatry 30, 320-321.
Abrams, R. and C. Vedak (1991). Prediction of ECT response in
melancholia. Convulsive Therapy 7, 81-84.
Agbayewa, M. O. (1986). Earlier psychiatric morbidity in patients with
Alzheimer's disease. Journal of the American Geriatrics Society 34,
561-564.
Agha, M. and M. Ibrahim (1984). Algorithm AS203 maximum likelihood
estimates of mixtures of distributions. Applied Statistics 33, 327-332.
Albert, M. L. (1978). Subcortical dementia. In R. Katzman, R. D. Terry,
and K. L. Birch (Eds.), Alzheimer's Disease: Senile Dementia and
Related Disorders, pp. 173-180. New York: Raven Press.
Albert, M. L., R. G. Feldman and A. L. Willis (1974). The "subcortical
dementia" of progressive supranuclear palsy. Journal of Neurology,
Neurosurgery and Psychiatry 37, 121-130.
Alexander, G. E., M. D. Crutcher and M. R. DeLong (1990). Basal
ganglia-thalamocortical circuits: parallel substrates for motor,
oculomotor, "prefrontal" and "limbic" functions. Progress in Brain
Research 85, 119-146.
Alexander, G. E. and M. R. DeLong (1985). Microstimulation of the primate
neostriatum: II. Somatotopic organization of striatal microexcitable
zones and their relation to neuronal response properties. Journal of
Neurophysiology 53, 1417-1430.
Alexander, G. E., M. R. DeLong and P. L. Strick (1986). Parallel
organisation of functionally segregated circuits linking basal ganglia and
cortex. Annual Review of Neuro science 9, 357-381.
Alexopoulos, G. S., B. S. Meyers, R. C. Young, S. Mattis and T. Kakuma
(1993a). The course of geriatric depression with "reversible dementia":
a controlled study. American Journal of Psychiatry 150, 1693-1699.
Alexopoulos, G. S., R. C. Young, J. A. Haycox, C. A. Shamoian and J. P.

292
References 293

Blass (1985). Dexsamethosone suppression test independence with


reversible dementia. Psychiatry Research 16, 277-285.
Alexopoulos, G. S., R. C. Young, K. W. Lieberman and C. A. Shamoian
(1987). Platelet MAO activity in geriatric patients with depression and
dementia. American Journal of Psychiatry 144, 1480-1483.
Alexopoulos, G. S., R. C. Young and B. S. Meyers (1993b). Geriatric
depression: age of onset and dementia. Biological Psychiatry 34,
141-145.
Alexopoulos, G. S., R. C. Young, B. S. Meyers, R. C. Abrams and C. A.
Shamoian (1988). Late-onset depression (review). Psychiatric Clinics of
North America 11, 101-115.
Almkvist, O., L. O. Whulund, G. Andersson-Lundman, H. Bassun and
L. Backman (1992). White matter hyperintensity and
neuropsychological functions in dementia and healthy aging. Archives of
Neurology 49, 626-632.
American Psychiatric Association (1968). Diagnostic and Statistical Manual
of Mental Disorders (2nd ed.) (DSM-II). Washington, DC: American
Psychiatric Association.
American Psychiatric Association (1980). Diagnostic and Statistical Manual
of Mental Disorders (3rd ed.) (DSM-III). Washington, DC:
American Psychiatric Association.
American Psychiatric Association (1987). Diagnostic and statistical manual
of mental disorders (3rd ed. Revised) (DSM-III-R). Washington, DC:
American Psychiatric Association.
American Psychiatric Association (1994). Diagnostic and Statistical Manual
of Mental Disorders (4th ed.) (DSM-IV). Washington, DC: American
Psychiatric Association.
Amouyel, P., T. Brousseau, J. C. Pruchart and J. Dallongeville (1993).
Apolipoprotein E-e4 allele and Alzheimer's disease [letter]. Lancet 342,
1309.
Ananth, J., R. Gamal, M. Miller, M. Wohl and S. Vandewater (1993). Is the
routine CT head scan justified for psychiatric patients? A prospective
study. Journal of Psychiatry and Neuroscience 18, 69-73.
Anderson, J., E. Aabro, N. Gulmann, A. Hijelmsted and H. E. Pedersen
(1980). Antidepressive treatment in Parkinson's disease. Ada
Neurologica Scandinavica 62, 210-219.
Andreasen, N. C. (1980). Concepts, diagnosis and classification. In E. Paykel
(Ed.), Handbook of Affective Disorders. Edinburgh: Churchill
Livingstone.
Andreasen, N. C. (1988). Brain imaging: applications in psychiatry (review).
Science 239, 1381-1388.
Andreasen, N. C , W. M. Grove and R. Maurer (1980). Cluster analysis and
the classification of depression. British Journal of Psychiatry 137,
256-265.
Andrews, G., M. D. Neilson, C. Hunt, G. W. Stewart and L. G. Kiloh
(1990a). Diagnosis and personality of the long-term outcome of
depression. British Journal of Psychiatry 157, 13-18.
Andrews, G., G. Stewart, A. Morris-Yates, P. Holt and S. Henderson
(1990b). Evidence for a general neurotic syndrome. British Journal of
Psychiatry 157, 6-12.
Angst, J. (1966). Zur Aetiologie und Nosologie endogener depressiver
294 References
Psychosen. Berlin: Springer.
Antleman, S. and L. A. Chiodo (1981). Dopamine autoreceptor
sub-sensitivity: a mechanism common to the treatment of depression
and the induction of amphetamine psychosis. Biological Psychiatry 16,
717-727.
Arana, G. W., R. J. Workman and R. J. Baldessarini (1984). Association
between low plasma levels of dexamethasone and elevated levels of
cortisol in psychiatric patients given dexamethasone. American Journal
of Psychiatry 141, 1619-1620.
Austin, M.-R, N. Dougall, M. Ross, C. Murray, R. E. O'Carroll, A. Moffoot,
K. P. Ebmeier and G. M. Goodwin (1992a). Single photon emission
tomography with " m Tc-exametazime in major depression and the
pattern of brain activity underlying the psychotic/neurotic continuum.
Journal of Affective Disorders 26, 31-44.
Austin, M.-P. and P. Mitchell (1995). The anatomy of melancholia: does
fronto-subcortical pathophysiology underpin its psychomotor and
cognitive manifestations? Psychological Medicine (in press).
Austin, M.-P., M. Ross, C. Murray, R. E. O'Carroll, K. P. Ebmeier and
G. M. Goodwin (1992b). Cognitive function in major depression.
Journal of Affective Disorders 25, 21-30.
Awad, I. A., P. C. Johnson and R. F. Spetzler (1986). Incidental subcortical
lesions identified on magnetic resonance imaging in the elderly: II. Post
mortem pathological correlations. Stroke 17, 1090-1097.
Baddeley, A. and G. J. Hitch (1974). Working memory. In G. Bower (Ed.),
Recent Advances in Learning and Motivation, pp. 47-90. New York:
Academic Press.
Baldwin, R. C. (1993). Late-life depression and structural brain changes. A
review of recent magnetic resonance imaging research. International
Journal of Geriatric Psychiatry 8, 115-123.
Baldwin, R. C. and D. J. Jolley (1986). The prognosis of depression. British
Journal of Psychiatry 149, 574-583.
Barrett, J. E., J. A. Barrett, T. E. Oxman and P. D. Gerber (1988). The
prevalence of psychiatric disorders in a primary care practice. Archives
of General Psychiatry 45, 1100-1106.
Bartholomew, D. J. (1987). Latent Variable Models and Factor Analysis.
London: Charles Griffin.
Baxter, L. R., M. E. Phelps, J. C. Mazziotta, B. H. Guze, C. E. Selin, R. H.
Gerner and R. M. Sumida (1989). Reduction of prefrontal cortex
glucose metabolism common to three types of depression. Archives of
General Psychiatry 46, 243-250.
Baxter, L. R., M. E. Phelps, J* C. Mazziotta, J. M. Schwartz, R. H. Gerner,
C. E. Selin and R. M. Sumida (1985). Cerebral metabolic rates for
glucose in mood disorders. Archives of General Psychiatry 42, 441-447.
Beats, B. and R. Levy (1992). Imaging and affective disorder in the elderly.
Clinics in Geriatric Medicine 8, 267-274.
Bech, P., P. Allerup, L. F. Gram, P. Kragh-Sorensen, O. J. Rafaelsen,
N. Reisby, P. Vestergaard and the Danish University Antidepressant
Group (DUAG) (1988). The Diagnostic Melancholia Scale (DMS):
dimensions of endogenous and reactive depression with relationship to
the Newcastle scale. Journal of Affective Disorders 14, 161-170.
References 295
Beck, A. T., C. H. Ward, M. Mendelson, J. E. Mock and J. K. Erbaugh
(1961). An inventory for measuring depression. Archives of General
Psychiatry 4, 561-571.
Beigel, A. and D. L. Murphy (1971). Unipolar and bipolar affective illness:
differences in clinical characteristics accompanying depression. Archives
of General Psychiatry 24, 215-220.
Bench, C. J., R. J. Dolan, K. J. Priston and R. S. J. Prackowiak (1993a).
State and trait abnormalities of regional cerebral blood flow in
depression determined by positron emission tomography. Journal of
Cerebral Blood Flow and Metabolism 13 (Suppl 1), S503.
Bench, C. J., K. J. Friston, R. G. Brown, R. S. J. Prackowiak and R. J.
Dolan (1993b). Regional cerebral blood flow in depression measured by
positron emission tomography: the relationship with clinical dimensions.
Psychological Medicine 23, 579-590.
Bench, C. J., K. J. Priston, R. G. Brown, L. C. Scott, R. S. J. Prackowiak
and R. J. Dolan (1992). The anatomy of melancholia — focal
abnormalities of cerebral blood flow in major depression. Psychological
Medicine 22, 606-615.
Berent, S., B. Giordani, S. Lehtinen, D. Markel, J. B. Penney, H. A. Buchtel,
S. Starosta-Rubinstein, R. Hichwa and A. B. Young (1988). Positron
emission tomography scan investigation of Huntington's disease:
cerebral metabolic correlations of cognitive function. Annals of
Neurology 23, 541-546.
Berger, M., K.-M. Pirke, P. Doerr, J.-C. Krieg and D. von Zerssen (1984).
The limited utility of the dexamethasone suppression test for the
diagnostic process in psychiatry. British Journal of Psychiatry 145,
372-382.
Berman, K. F. Doran, A. R., D. Pickar and D. R. Weinberger (1993). Is the
mechanism of prefrontal hypofunction in depression the same as in
schizophrenia? Regional cerebral blood flow during cognitive activation.
British Journal of Psychiatry 162, 183-192.
Berrios, G. (1985). The psychopathology of affectivity: conceptual and
historical aspects. Psychological Medicine 15, 745-758.
Berrios, G. E. (1988). Melancholia and depression during the 19th century.
British Journal of Psychiatry 153, 298-304.
Blackburn, I. M. (1975). Mental and psychomotor speed in depression and
mania. British Journal of Psychiatry 126, 329-335.
Blazer, D., J. R. Bachar and D. C. Hughes (1987). Major depression with
melancholia: a comparison of middle-aged and elderly adults. Journal of
the American Geriatrics Society 35, 927-932.
Blazer, D., M. Swartz, M. Woodbury, K. G. Manton, D. Hughes and L. K.
George (1988). Depressive symptoms and depressive diagnoses in a
community population: use of a new procedure for analysis of
psychiatric classification. Archives of General Psychiatry 45, 1078-1084.
Blazer, D. G., R. C. Kessler, K. A. McGonagle and M. S. Swartz (1994).
The prevalence and distribution of major depression in a national
community sample: The National Comorbidity Survey. American
Journal of Psychiatry 151, 979-986.
Blumenthal, M. D. (1971). Heterogeneity and research on depressive
disorders. Archives of General Psychiatry 24, 524-531.
Bowen, D. M., A. Najlerahim, A. W. Procter, P. T. Francis and E. Murphy
296 References

(1989). Circumscribed changes of the cerebral cortex in neuropsychiatric


disorders. Proceedings of the National Academy of Sciences of the
United States of America 86, 9504-9508.
Boyce, P. and I. Hickie (1994). A brave new world in managing depression
— or is it? Australian Family Physician 23, 627-632.
Boyd, J. H., J. D. Burke, E. Gruenberg, C. E. Holzer, D. S. Rae, L. K.
George, M. Karno, R. Stoltzman, M. A. McEvoy and G. Nestadt (1984).
Exclusion criteria of DSM-III: a study of co-occurrence of hierarchy-free
syndromes. Archives of General Psychiatry 41, 983-989.
Boyd, J. L., C. A. Cruickshank, C. W. Kenn, P. Madeley, R. H. S. Mindham,
A. G. Oswald, R. J. Smith and E. G. S. Spokes (1991). Cognitive
impairment and dementia in Parkinson's disease: a controlled study.
Psychological Medicine 21, 911-921.
Brand, A. N., J. Jolles and C. Gispen-de Wied (1992). Recall and recognition
memory deficits in depression. Journal of Affective Disorders 25, 77-86.
Brand, J. and N. Butters (1986). The neuropsychology of Huntington's
disease. Trends in Neuroscience 9, 118-120.
Brandon, S., P. Cowley and C. McDonald (1984). Electroconvulsive therapy:
results in depressive illness from the Leicestershire trial. British Medical
Journal 288, 22-25.
Brockington, I. F., E. Altman, V. Hillier, H. Y. Meltzer and S. Nand (1982).
The clinical picture of bipolar affective disorder in its depressed phase: a
report from London and Chicago. British Journal of Psychiatry 141,
558-562.
Brodaty, H., P. Boyce, K. Wilhelm, P. Mitchell and G. Parker (1987). The
establishment of a mood disorders unit. Australian and New Zealand
Journal of Psychiatry 21, 375-381.
Brodaty, H., K. Peters, P. Boyce, I. Hickie, G. Parker, P. Mitchell and
K. Wilhelm (1991). Age and depression. Journal of Affective
Disorders 23, 137-149.
Brown, A. S. and S. Gershon (1993). Dopamine and depression. Journal of
Neural Transmission 91, 75-109.
Brown, F. W., R. J. Lewine, P. A. Hudgins and S. C. Risch (1992). White
matter hyperintensity signals in psychiatric and nonpsychiatric subjects.
American Journal of Psychiatry 149, 620-625.
Brown, R. P., P. M. Stoll, P. E. Stokes, A. Frances, J. Sweeney, J. H. Kocsis
and J. J. Mann (1988). Adrenocortical hyperactivity in depression:
effects of agitation, delusions, melancholia, and other illness variables.
Psychiatry Research 23, 167-178.
Brown, R. P., J. Sweeney, E. Loutsch, E. Kocsis, J. Kocsis and A. Frances
(1984). Involutional melancholia revisited. American Journal of
Psychiatry 141, 24-28.
Brozoski, T. J., R. M. Brown, H. E. Rosvold and P. S. Goldman (1979).
Cognitive deficit caused by regional depletion of dopamine in prefrontal
cortex of rhesus monkey. Science 205, 929-932.
Bruce, M. L. and P. J. Leaf (1989). Psychiatric disorders and 15-month
mortality in a community sample of older adults. American Journal of
Public Health 79, 727-730.
Buchan, H., E. Johnstone, K. McPherson, R. Palmer, T. Crow and
S. Brandon (1992). Who benefits from electroconvulsive therapy?
Combined results of the Leicester and Norwich Park trials. British
References 297
Journal of Psychiatry 160, 355-359.
Buchsbaum, M. S., J. Wu, L. E. DeLisi, H. Holcomb, R. Kessler, J. Johnson,
A. C. King, E. Hazlett, K. Langston and R. M. Post (1986). Frontal
cortex and basal ganglia metabolic rates assessed by positron emission
tomography with [18F]2-deoxyglucose in affective illness. Journal of
Affective Disorders 10, 137-152.
Burns, A., R. Jacoby and R. Levy (1990). Psychiatric phenomena in
Alzheimer's disease: III. Disorders of mood. British Journal of
Psychiatry 157, 81-86.
Burvill, P. (1994). Psychiatric aspects of cerebro-vascular disease. In E. Chiu
and D. Ames (Eds.), Functional Psychiatric Disorders of the Elderly,
pp. 407-424. Cambridge: Cambridge University Press.
Burvill, P. W. and W. D. Hall (1994). Predictors of increased mortality in
elderly depressed patients. International Journal of Geriatric
Psychiatry 9, 219-227.
Burvill, P. W., W. D. Hall, H. G. Stampfer and J. P. Emmerson (1989). A
comparison of early-onset and late-onset depressive illness in the elderly.
British Journal of Psychiatry 155, 673-679.
Burvill, P. W., W. D. Hall, H. G. Stampfer and J. P. Emmerson (1991). The
prognosis of depression in old age. British Journal of Psychiatry 158,
64-71.
Burvill, P. W., H. Stampfer and W. Hall (1986). Review: does depressive
illness in the elderly have a poor prognosis? Australian and New
Zealand Journal of Psychiatry 20, 422-427.
Butters, N., D. Sax, K. Montgomery and S. Tarlow (1978). Comparison of
the neuropsychological deficits associated with early and advanced
Huntington's disease. Archives of Neurology 35, 585-589.
Butters, N., J. Wolfe, M. Martone, E. Granholme and L. S. Cermak (1985).
Memory disorders associated with Huntington's disease: verbal recall,
verbal recognition and procedural memory. Neuropsychologia 23,
729-743.
Caine, E. D., M. H. Ebert and H. Weingartner (1977). An outline for the
analysis of dementia: the memory disorder of Huntington's disease.
Neurology 27, 1087-1092.
Calloway, S. P., R. J. Dolan, P. Fonagy, V. F. A. De Souza and A. Wakeling
(1984). Endocrine changes and clinical profiles in depression: I. The
dexamethasone suppression test. Psychological Medicine 14, 749-758.
Carney, M. W. P., M. Roth and R. F. Garside (1965). The diagnosis of
depressive syndromes and the prediction of E.C.T. response. British
Journal of Psychiatry 111, 659-674.
Carpenter, W. T., R. W. Buchanan, B. Kirkpatrick, C. Tamminga and
F. Wood (1993). Strong inference, theory testing, and the neuroanatomy
of schizophrenia. Archives of General Psychiatry 50, 825-831.
Carroll, B. J. (1989). Diagnostic validity and laboratory studies: rules of the
game. In L. N. Robins and J. E. Barrett (Eds.), The Validity of
Psychiatric Diagnosis, pp. 229-245. New York: Raven Press.
Carroll, B. J. (1991). Psychopathology and neurobiology of manic-depressive
disorders. In B. J. Carroll and J. E. Barrett (Eds.), Psychopathology and
the Brain, pp. 265-285. New York: Raven Press.
Carroll, B. J. (1994). Brain mechanisms in manic depression. Clinical
Chemistry 40, 303-308.
298 References
Carroll, B. J., M. Feinberg, J. F. Greden, J. Tarika, A. A. Albala, R. F.
Haskett, N. M. James, Z. Kronfol, N. Lohr, M. Steiner, J. P. de Vigne
and E. Young (1981a). A specific laboratory test for the diagnosis of
melancholia: standardization, validation and clinical utility. Archives of
General Psychiatry 38, 15-22.
Carroll, B. J., M. Feinberg, P. Smouse, S. Rawson and J. Greden (1981b).
The Carroll rating scale for depression: 1. Development, reliability, and
validation. British Journal of Psychiatry 138, 194-200.
Cassens, G., L. Wolfe and M. Zola (1990). The neuropsychology of
depressions. Neuropsychology Update Series 2, 202-212.
Channon, S., J. E. Baker and M. M. Robertson (1993). Working memory in
clinical depression: an experimental study. Psychological Medicine 23,
87-91.
Charney, D. S. and J. C. Nelson (1981). Delusional and nondelusional
unipolar depression: further evidence for distinct subtypes. American
Journal of Psychiatry 138, 328-333.
Christensen, P., L. F. Gram, P. Kragh-Sorensen and S. Nielsen (1986).
Afternoon cortisol levels before (spontaneous) and after suppression
with dexamethasone or oxazepam in depressed patients. Journal of
Affective Disorders 10, 171-176.
Cloitre, M., M. M. Katz and H. M. van Praag (1993). Psychomotor
agitation and retardation. In C. Costello (Ed.), Symptoms of
Depression, pp. 207-226. New York: John Wiley.
Coffey, C. E., G. S. Figiel, W. R. Djang, W. B. Saunders and R. D. Weiner
(1989). White matter hyperintenstiy on magnetic resonance imaging:
clinical and neuroanatomic correlates in the depressed elderly. Journal
of Neuropsychiatry and Clinical Neuroscience 1, 135-144.
Coffey, C. E., G. S. Figiel, W. T. Djang, M. Cress, W. B. Saunders and
R. D. Weiner (1988). Leukoencephalopathy in elderly depressed patients
referred for ECT. Biological Psychiatry 24, 143-161.
Coffey, C. E., G. S. Figiel, W. T. Djang and R. D. Weiner (1990).
Subcortical hyperintensity on magnetic resonance imaging: a
comparison of normal and depressed elderly subjects. American Journal
of Psychiatry 147, 187-189.
Coffey, C. E., R. D. Weiner, W. T. Djang, G. S. Figiel, S. A. R. Soady, L. J.
Patterson, P. D. Holt, C. E. Spritzer and W. E. Wilkinson (1991). Brain
anatomic effects of electroconvulsive therapy. Archives of General
Psychiatry 48, 1013-1021.
Coffey, C. E., W. E. Wilkinson, I. A. Parashos, S. A. Soady, R. J. Sullivan,
L. J. Patterson, G. S. Figiel, M. C. Webb, C. E. Spritzer and W. T.
Djang (1992). Quantitative cerebral anatomy of the aging human brain:
a cross-sectional study using magnetic resonance imaging.
Neurology 42, 527-536.
Coffey, C. E., W. E. Wilkinson, R. D. Weiner, I. A. Parashos, W. T. Djang,
M. C. Webb, G. S. Figiel and C. E. Spritzer (1993). Quantitative
cerebral anatomy in depression: a controlled magnetic resonance
imaging study. Archives of General Psychiatry 50, 7-16.
Cohen, R. M., W. E. Semple, M. Gross, T. E. Nordahl, A. C. King,
D. Pickar and R. M. Post (1989). Evidence for common alterations in
cerebral glucose metabolism in major affective disorders and
schizophrenia. Neuropsychopharmacology 2, 241-253.
References 299

Cohen, R. M., H. Weingartner, S. A. Smallberg, D. Pickar and D. L. Murphy


(1982). Effort and cognition in depression. Archives of General
Psychiatry 39, 593-598.
Conwell, Y., J. C. Nelson, K. M. Kim and C. M. Mazure (1989). Depression
in late life: age of onset as marker of a subtype. Journal of Affective
Disorders 17, 189-195.
Cooper, J. A., H. J. Sagar, N. Jordan, N. S. Harvey and E. V. Sullivan
(1991). Cognitive impairment in early, untreated Parkinson's disease
and its relationship to motor disability. Brain 114, 2095-2122.
Copeland, J. R. M. (1988). Physical ill-health, age and depression. In
T. Helgason and R. J. Daly (Eds.), Depressive Illness: Prediction of
Course and Outcome, pp. 118-225. Berlin: Springer-Verlag.
Coppen, A., M. Abou-Saleh, P. Milln, M. Metcalfe, J. Harwood and
J. Bailey (1983). Dexamethasone suppression test in depression and
other psychiatric illness. British Journal of Psychiatry 142, 498-504.
Corder, B., A. M. Saunders, W. J. Strittmatter, D. E. Schmechel, P. C.
Gaskell and G. N. Small (1993). Gene dose of apolipoprotein E-e4 allele
and the risk of Alzheimer's disease in late onset families. Science 261,
921-923.
Cornell, D. G., R. Suarez and S. Berent (1984). Psychomotor retardation in
melancholic and non-melancholic depression: cognitive and motor
components. Journal of Abnormal Psychology 932, 150—157.
Coryell, W., G. Winokur, T. Shea, J. D. Maser, J. Endicott and H. S.
Akiskal (1994). The long-term stability of depressive subtypes.
American Journal of Psychiatry 151, 199-204.
Costello, C. (1993). From symptoms of depression to syndromes of
depression. In C. Costello (Ed.), Symptoms of Depression, pp. 291-302.
New York: John Wiley.
Creasey, H. and S. I. Rapoport (1985). The aging human brain. Annals of
Neurology 17, 2-10.
Cronholm, B. and J. Ottosson (1961). Memory function in endogenous
depression. Archives of General Psychiatry 5, 193-197.
Cummings, J. L. (1992). Depression and Parkinson's disease: a review.
American Journal of Psychiatry 149, 443-454.
Cummings, J. L. (1993). The neuroanatomy of depression. Journal of
Clinical Psychiatry 54 (Suppl. 11), 14-20.
Cummings, J. L., B. Miller, M. A. Hill and R. Neshkes (1987).
Neuropsychiatric aspects of multi-infarct dementia and dementia of the
Alzheimer type. Archives of Neurology 44, 389-393.
Curran, D. (1937). The differentiation of neuroses and manic-depressive
psychoses. Journal of Mental Science 86, 156-191.
Curran, D. and W. P. Mallinson (1941). Depressive states in war. British
Medical Journal 1, 305-309.
Curran, S. M., C. M. Murray, M. Van Beck, N. Dougall, R. E. O'Carroll,
M.-P. Austin, K. P. Ebmeier and G. M. Goodwin (1993). A single
photon emission computerised tomography study of regional brain
function in elderly patients with major depression and with
Alzheimer's-type dementia. British Journal of Psychiatry 163, 155-165.
D'Amato, R. J., R. M. Zweig, P. J. Whitehouse, G. L. Wenk, H. S. Singer,
R. Mayeux, D. L. Price and D. H. Snyder (1987). Aminergic systems in
Alzheimer's disease and Parkinson's disease. Annals of Neurology 22,
300 References

229-236.
Davidson, J., S. Lipper, W. W. K. Zung, R. Strickland, R. Krishnan and
S. Mahorney (1984). Validation of four definitions of melancholia by the
dexamethasone suppression test. American Journal of Psychiatry 141,
1220-1223.
Davis, J. M., S. H. Koslow, R. D. Gibbons, J. W. Maas, C. L. Bowden,
R. Casper, I. Hanin, J. I. Javaid, S. S. Chang and P. E. Stokes (1988).
Cerebrospinal fluid and urinary biogenic amines in depressed patients
and healthy controls. Archives of General Psychiatry 45, 705-717.
Davis, K. L., R. S. Kahn, G. Ko and M. Davidson (1985). Dopamine in
schizophrenia: a review and reconceptualization. American Journal of
Psychiatry 148, 1474-1486.
Davis, P. C , S. S. Mirra and N. Alazraki (1994). The brain in older persons
with and without dementia: findings on MR, PET and SPECT images.
American Journal of Radiology 162, 1267-1278.
De Keyser, J., P. Herregodts and G. Ebinger (1990). The mesoneocortical
dopamine neuron system (Review). Neurology 40, 1660-1662.
Delvenne, V., F. Delecluse, P. P. Hubain, A. Schoutens, V. De Maertelaer
and J. Mendlewicz (1990). Regional cerebral blood flow in patients with
affective disorders. British Journal of Psychiatry 157, 359-365.
Derogatis, L. and P. Cleary (1977). Confirmation of the dimensional
structure of the SCL-90: a study in construct validation. Journal of
Clinical Psychology 33, 981-989.
Detre, T., H. Himmelhoch, M. Swartzburg, C. M. Anderson, R. Byck and
D. J. Kupfer (1972). Hypersomnia and manic-depressive disease.
American Journal of Psychiatry 128, 1303-1305.
Devanand, D. P., H. A. Sackeim, E.-S. Lo, T. Cooper, G. Huttinot, J. Prudic
and F. Ross (1991). Serial dexamethasone suppression tests and plasma
dexamethasone levels. Archives of General Psychiatry 48, 525-533.
Divac, I., F. Fonnum and J. Storm-Mathisen (1977). High affinity uptake of
glutamate in terminals of corticostriatal axons. Nature 266, 377-378.
Dolan, R. J., C. J. Bench, P. F. Liddle, K. J. Friston, C. D. Frith, P. M.
Grasby and R. S. J. Frackowiak (1993). Dorsolateral prefrontal cortex
dysfunction in the major psychoses: symptom or disease specificity?
Journal of Neurology, Neurosurgery and Psychiatry 56, 1290-1294.
Dolan, R. J., A. M. Poynton, P. K. Bridges and M. R. Trimble (1990).
Altered magnetic resonance white-matter Tl values in patients with
affective disorder. British Journal of Psychiatry 157, 107-110.
Donnelly, E. F., D. L. Murphy, F. K. Goodwin and I. N. Waldman (1982).
Intellectual function in primary affective disorders. British Journal of
Psychiatry 140, 633-636.
Dooneief, G., E. Mirabello, K. Bell, K. Marder, Y. Stern and R. Mayeux
(1992). An estimate of the incidence of depression in idiopathic
Parkinson's disease. Archives of Neurology 49, 305-307.
Drayer, B. P. (1988). Imaging of the aging brain: Part I. Normal findings.
Radiology 166, 785-796.
Drevets, W. C , T. O. Videen, J. L. Price, S. H. Preskorn, T. Carmichael and
M. E. Raichle (1992). A functional anatomical study of unipolar
depression. Journal of Neuroscience 12, 3628-3641.
Duggan, C. F., A. S. Lee and R. M. Murray (1991). Do different subtypes of
hospitalized depressives have different long-term outcomes? Archives of
References 301

General Psychiatry 48, 308-312.


Dunner, D. L., T. Dwyer and R. R. Fieve (1976). Depressive symptoms in
patients with unipolar and bipolar affective disorder. Comprehensive
Psychiatry 17, 447-451.
Dupont, R. M., T. L. Jernigan, N. Butters, D. Delis, J. R. Hesselink,
W. Heindel and J. C. Gillin (1990). Subcortical abnormalities detected
in bipolar affective disorder using magnetic resonance imaging. Archives
of General Psychiatry 47, 55-59.
Eagles, J. M. and L. J. Whalley (1985). Ageing and affective disorders: the
age at first onset of affective disorders in Scotland 1969-1978. British
Journal of Psychiatry 147, 180-187.
Ebrahim, S. and F. Nouri (1987). Affective illness after stroke. British
Journal of Psychiatry 151, 52-56.
Elkin, I., T. Shea, J. T. Watkins, S. D. Imber, S. M. Sotsky, J. F. Collins,
D. R. Glass, P. A. Pilkonis, W. R. Leber, J. P. Docherty, S. J. Fiester
and M. B. Parloff (1989). National Institute of Mental Health
Treatment of Depression Collaborative Research Program. Archives of
General Psychiatry 46, 971-982.
Ellard, J. (1985). The music of psychiatry. Australian and New Zealand
Journal of Psychiatry 19, 153-156.
Emery, V. O. and T. E. Oxman (1992). Update on the dementia spectrum of
depression. American Journal of Psychiatry 149, 305-317.
Endicott, J. and L. Spitzer (1978). A diagnostic interview: the schedule for
affective disorders and schizophrenia. Archives of General
Psychiatry 35, 837-844.
Esiri, M. N. (1994). Dementia and normal ageing: neuropathology. In F. A.
Huppert, C. Brayne, and D. W. O'Connor (Eds.), Dementia and
Normal Ageing, pp. 385-436. Cambridge: Cambridge University Press.
Evarts, E. V., H. Teravainen and D. B. Calne (1981). Reaction time in
Parkinson's disease. Brain 104, 167-186.
Everitt, B. (1981). A Monte Carlo investigation of the likelihood ratio test
for the number of components in a mixture of normal distribution.
Multivariate Behavioral Research 16, 171-180.
Eysenck, H. J. (1970). The classification of depressive illness. British Journal
of Psychiatry 117, 241-250.
Fahn, S. and R. L. Elton (1987). UPDRS Development Committee: Unified
Parkinson's Disease Rating Scale. In S. Fahn, C. D. Marsden, D. Calne,
and M. Goldstein (Eds.), Recent Developments in Parkinson's Disease.
Volume 2, pp. 293-304. Florham Park, NY: MacMillan Healthcare
Information.
Fairchild, C. J., A. J. Rush, N. Vasavada, D. E. Giles and M. Khatami
(1986). Which depressions respond to placebo? Psychiatric
Research 18, 217-226.
Farmer, A. and P. McGuffin (1989). The classification of the depressions:
contemporary confusion revisited. British Journal of Psychiatry 155,
437-443.
Fazekas, F. (1989). Magnetic resonance signal abnormalities in
asymptomatic individuals: their incidence and functional correlates.
European Journal of Neurology 29, 164-168.
Fedoroff, J. P., S. E. Starkstein, A. W. Forrester, F. H. Geisler, R. E. Jorge,
S. V. Arndt and R. G. Robinson (1992). Depression in patients with
302 References

acute traumatic brain injury. American Journal of Psychiatry 149,


918-923.
Ferrier, I. N. and F. K. Perry (1992). Post-mortem studies in affective
disorder. Psychological Medicine 22, 835-838.
Figiel, G., C. E. Coffey, W. T. Djang, G. Hoffman and P. M. Doraiswamy
(1990a). Brain magnetic resonance imaging findings in ECT-induced
delirium. Journal of Neuropsychiatry and Clinical Neuroscience 2,
53-58.
Figiel, G. S., K. R. R. Krishnan and P. M. Doraiswamy (1990b). Subcortical
structural changes in ECT-induced delirium. Journal of Geriatric
Psychiatry and Neurology 3, 172-176.
Figiel, G. S., K. R. R. Krishnan, P. M. Doraiswamy, V. P. Rao, C. B.
Nemeroff and O. B. Boyko (1991). Subcortical hyperintensities on brain
magnetic resonance imaging: a comparison between late age onset and
early onset elderly depressed subjects. Neurobiology of Aging 26,
245-247.
Figiel, G. S., R. Krishnan, J. C. Breitner and C. B. Nemeroff (1989).
Radiologic correlates of antidepressent-induced delirium: the possible
significance of basal ganglia lesions. Journal of Neuropsychiatry and
Clinical Neurosciences 1, 188-190.
Fleiss, J. L., B. J. Gurland and J. E. Cooper (1971). Some contributions to
the development of psychopathology. British Journal of Psychiatry 119,
647-656.
Fleminger, S. (1992). Control of simultaneous movements distinguishes
depressive motor retardation from Parkinson's disease and neuroleptic
parkinsonism. Brain 115, 1459-1480.
Flowers, K. A. and C. Robinson (1985). The effect of Parkinson's disease on
the ability to maintain a mental test set. Journal of Neurology,
Neurosurgery and Psychiatry 48, 517-529.
Folstein, S. E. (1989). Huntingtons's Disease: a Disease of Families.
Baltimore: Johns Hopkins University Press.
Friedman, A. S. (1964). Minimal effects of severe depression on cognitive
functioning. Journal of Abnormal and Social Psychology 69, 237-243.
Friston, K. J., C. D. Frith, P. F. Liddle and R. S. J. Frackowiak (1991).
Comparing functional (PET) images: the assessment of significant
change. Journal of Cerebral Blood Flow and Metabolism 11, 690-699.
Gangestad, S. and M. Snyder (1985). "To carve nature at its joints": on the
existence of discrete classes in personality. Psychological Review 92,
317-349.
Garside, R. F., D. W. K. Kay, I. C. Wilson, I. D. Deaton and M. Roth
(1971). Depressive symptoms and the classification of patients.
Psychological Medicine 1, 333-338.
George, A. E., M. J. de Leon, A. Kalnin, L. Rosner, A. Goodgold and
N. Chase (1986). Leukoencephalopathy in normal and pathologic aging:
MRI of brain lucencies. American Journal of Neuroradiology 7, 567-570.
George, M. S., H. A. Ring, D. C. Costa, P. J. Ell, K. Kouris and P. H.
Jarritt (1991). Clinical use of SPET imaging in psychiatric and
neurological disease. In M. S. George, H. A. Ring, D. C. Costa, P. J. Ell,
K. Kouris, and P. H. Jarritt (Eds.), Neuroactivation and Neuroimaging
with SPET, pp. 51-120. London: Springer-Verlag.
Georgotas, A., P. E. Stokes, W. E. Hapworth, O. M. Kim, C. Farelli, P. M.
References 303
Stoll, E. Sinaiko and R. E. McCue (1987). The relationship of the
dexamethasone suppression test to subtypes of depression and to
symptomatic severity in the elderly. Journal of Affective Disorders 10,
51-57.
Gerfen, C. R. (1992). The neostriatal mosaic: multiple levels of
compartmental organization in the basal ganglia. Annual Review of
Neuroscience 15, 285-320.
Gillespie, R. D. (1929). The clinical differentiation of types of depression.
Guy's Hospital Report 2, 306-344.
Giordani, B., M. J. Boivin, S. Berent, A. T. Betley, R. A. Koeppe, J. M.
Rothley, J. G. Modell, R. D. Hichwa and D. E. Kuhl (1990). Anxiety
and cerebral cortical metabolism in normal persons. Psychiatry
Research: Neuroimaging 35, 49-60.
Glassman, A. H. and S. P. Roose (1981). Delusional depression: a distinct
clinical entity. Archives of General Psychiatry 38, 424-427.
Goetz, C. G. (1992). The historical background of behavioral studies in
Parkinson's disease. In S. J. Huber and J. L. Cummings (Eds.),
Parkinson's Disease: Neurobehavioral Aspects, pp. 3-9. New York:
Oxford University Press.
Golden, R. (1982). A taxometric model for the detection of a conjectured
latent taxon. Multivariate Behavioral Research 17, 389-416.
Golden, R. R. (1991). Bootstrapping taxometrics: on the development of a
method for detection of a single major gene. In W. M. Grove and
D. Cicchetti (Eds.), Thinking Clearly about Psychology. Volume 2:
Personality and Psychopathology, pp. 259-294. Minneapolis: University
of Minnesota Press.
Goldman-Rakic, P. S. (1991). Prefrontal cortical dysfunction in
schizophrenia: the revelance of working memory. In B. J. Carrol and
J. E. Barrett (Eds.), Psychopathology and the Brain, pp. 1-23. New
York: Raven Press.
Goldsamt, L. A., M. M. Katz, P. Stokes, G. Parker and A. Miller (1995).
Cross-national stability of the CORE system for diagnosing melancholia,
pp. In submission.
Goodwin, F. K. and K. R. Jamison (1990). Manic-depressive Illness. New
York: Oxford University Press.
Goodwin, G. M., M.-P. Austin, N. Dougall, M. Ross, C. Murray, R. E.
O'Carroll, A. Moffoot, N. Prentice and K. P. Ebmeier (1993). State
changes in brain activity shown by the uptake of " m Tc-exametazime
with single photon emission tomography in major depression before and
after treatment. Journal of Affective Disorders 29, 243-253.
Gotham, A.-M., R. G. Brown and C. D. Marsden (1986). Depression in
Parkinson's disease: a quantitative and qualitative analysis. Journal of
Neurology, Neurosurgery and Psychiatry 49, 381-389.
Gotham, A. M., R. G. Brown and C. D. Marsden (1988). "Frontal"
cognitive function in patients with Parkinson's disease "on" and "off"
levodopa. Brain 111, 299-321.
Graybiel, A. M. and C. W. Ragsdale (1978). Histochemically distinct
compartments in the striatum of human being, monkey and cat
demonstrated by the acetylthiocholinesterase staining method.
Proceedings of the National Academy of Science USA 75, 5723-5726.
Grayson, D. (1986). Assessment of evidence for the categorical view of
304 References
schizophrenia [letter]. Archives of General Psychiatry 43, 712-714.
Grayson, D. A. (1987). Can categorical and dimensional views of psychiatric
illness be distinguished? British Journal of Psychiatry 151, 355-361.
Greden, J. F. and B. J. Carroll (1981). Psychomotor function in affective
disorders: an overview of new monitoring techniques. Americal Journal
of Psychiatry 138, 1441-1448.
Greenwald, B. S. and E. Kramer-Ginsberg (1988). Age at onset in geriatric
depression: relationship to clinical variables. Journal of Affective
Disorders 15, 61-68.
Grove, W. M. and N. C. Andreasen (1989). Quantitative and qualitative
distinctions between psychiatric disorders. In L. N. Robins and J. E.
Barrett (Eds.), The Validity of Psychiatric Diagnosis, pp. 127-141. New
York: Raven Press.
Grove, W. M. and A. Tellegen (1985). Distinguishing types from continua:
not as easy as it looks. Unpublished manuscript.
Grunwald, F., R. Horn, O. Rieker, E. Klemm, C. Menzel, H. J. Moller and
H. J. Biersack (1993). HMPAO-SPECT bei Demenz vom
Alzheimer-Typ und Major Depression mit mnestischen Storungen
[HMPAO SPECT in dementia of the Alzheimer type and major
depression with memory impairment]. Nuklearmedizin 32, 128-133.
Gunther, W., R. Gunther, P. Streck, H. Romig and A. Rodel (1988).
Psychomotor disturbances in psychiatric patients as a possible basis for
new attempts at differential diagnosis and therapy. European Archives
of Psychiatry and Neurological Science 237, 65-73.
Gur, R. C, R. J. Erwin and R. E. Gur (1992). Neurobehavioral probes for
physiologic neuroimaging studies. Archives of General Psychiatry 49,
409-414.
Gur, R. C , R. E. Gur, S. M. Resnick, B. E. Skolnick, A. Alavi and
M. Reivich (1987). The effect of anxiety on cortical cerebral blood flow
and metabolism. Journal of Cerebral Blood Flow and Metabolism 7,
173-177.
Gurland, B. J. (1976). The comparative frequency of depression in various
adult age groups. Journal of Gerontology 31, 283-292.
Guthrie, S. (1991). The impact of dexamethasone pharmacokinetics on the
DST: a review. Psychopharmacological Bulletin 27, 565-576.
Hadzi-Pavlovic, D., I. Hickie, H. Brodaty, P. Boyce, P. Mitchell, K. Wilhelm
and G. Parker (1993). Inter-rater reliability of a refined index of
melancholia: the CORE system. Journal of Affective Disorders 27,
155-162.
Hagman, J. O., M. S. Buchsbaum, J. C. Wu, S. J. Rao, C. A. Reynolds and
B. J. Blinder (1990). Comparison of regional brain metabolism in
bulimia nervosa and affective disorder assessed with positron emission
tomography. Journal of Affective Disorders 19, 153-162.
Halaris, A. E., K. T. Belendiuk and D. X. Freedman (1975). Antidepressant
drugs affect dopamine uptake. Biochemical Pharmacology 24,
1896-1897.
Hamilton, M. (1960). A rating scale for depression. Journal of Neurology,
Neurosurgery and Psychiatry 23, 56-62.
Hamilton, M. (1967). Development of a rating scale for primary depressive
illness. British Journal of Social and Clinical Psychology 6, 278-296.
Hamilton, M. and J. M. White (1959). Clinical syndromes in depressive
References 305

states. Journal of Mental Science 105, 985-998.


Harrell, L. E. Duvall, E., D. G. Folks, L. Duke, A. Bartolucci, T. Conboy,
R. Callaway and D. Kerns (1991). The relationship of high-intensity
signals on magnetic resonance images to cognitive and psychiatric state
in Alzheimer's disease. Archives of Neurology 48, 1136-1140.
Hart, B. L. (1988). Biological basis of the behaviour of sick animals.
Neuroscience and Behavioral Reviews 12, 123-137.
Hart, R. P. and J. A. Kwentus (1987). Psychomotor slowing and
subcortical-type dysfunction in depression. Journal of Neurology,
Neuro surgery and Psychiatry 50, 1263-1266.
Hasselbalch, S. G., G. Oberg, S. A. Sorensen, A. R. Andersen, G. Waldemar,
J. F. Schmidt, K. Fenger and O. B. Paulson (1992). Reduced regional
cerebral blood flow in Huntington's disease studied by SPECT. Journal
of Neurology, Neurosurgery and Psychiatry 55, 1018-1023.
Hasselblad, V. (1966). Estimation of parameters for a mixture of normal
distributions. Technometrics 8, 431—444.
Hayward, L., S. R. Zubrick and W. Hall (1985). Early sensory-perceptual
changes in Huntington's disease. Australian and New Zealand Journal of
Psychiatry 19, 384-389.
Healy, D. and J. M. G. Williams (1988). Dysrhythmia, dysphoria, and
depression: the interaction of learned helplessness and circadian
dysrhythmia in the pathogenesis of depression. Psychological
Bulletin 103, 163-178.
Henderson, A. S. (1994). Does ageing protect against depression? Social
Psychiatry and Epidemiology 29, 107-109.
Henderson, A. S., I. M. Montgomery and C. L. Williams (1972).
Psychological immunisation: a proposal for preventive psychiatry. The
Lancet 1, 111-1113.
Henderson, D. K. and R. D. Gillespie (1927). A Text-Book of Psychiatry for
Students and Practitioners. London: Oxford University Press.
Henderson, S., D. G. Byrne and P. Duncan-Jones (1981). Neurosis and the
Social Environment. London: Academic Press.
Henry, G. M., H. Weingartner and D. L. Murphy (1973). Influence of
affective states and psychoactive drugs on verbal learning and memory.
American Journal of Psychiatry 130, 966-971.
Hickie, I., B. Parsonage and G. Parker (1990). Prediction of response to
electroconvulsive therapy: preliminary validation of a sign-based
typology of depression. British Journal of Psychiatry 157, 65-71.
Hickie, I., E. Scott, P. Mitchell, K. Wilhelm, M.-P. Austin and B. Bennett
(1995). Subcortical hyperintensities on magnetic resonance imaging:
clinical correlates and prognostic significance in patients with severe
depression. Biological Psychiatry 37, 151-160.
Hobson, R. F. (1953). Prognostic factors in ECT. Journal of Neurology,
Neurosurgery and Psychiatry 16, 275-281.
Hoffman, G. M. A., J. C. Gonze and J. Mendlewicz (1985). Speech pause
time as a method for the evaluation of psychomotor retardation in
depressive illness. British Journal of Psychiatry 146, 535-538.
Holden, N. L. (1983). Depression and the Newcastle scale: their relationship
to the dexamethasone suppression test. British Journal of
Psychiatry 142, 505-507.
Holsboer, F., D. Haack, A. Gerken and P. Vecsei (1984). Plasma
306 References

dexamethasone concentrations and differential suppression response of


cortisol and corticosterone in depressives and controls. Biological
Psychiatry 19, 281-291.
Holsboer, F., K. Wiedemann and E. Boll (1986a). Shortened dexamethasone
half-life in depressed dexamethasone nonsuppressors. Archives of
General Psychiatry 43, 813-814.
Holsboer, F., K. Wiedemann, A. Gerken and E. Boll (1986b). The plasma
dexamethasone variable in depression: test-retest studies and early
biophase kinetics. Psychiatry Research 17, 97-103.
Honer, W. G., T. Hurwitz, D. K. Li, M. Palmer and D. W. Paty (1987).
Temporal lobe involvement in multiple sclerosis patients with
psychiatric disorders. Archives of Neurology 44, 187-190.
Hope, K. (1969). The classification of depressive illness [book review].
British Journal of Psychiatry 115, 731-741.
Hopkinson, G. A. (1964). A genetic study of affective illness in patients over
fifty. British Journal of Psychiatry 110, 244-254.
Huber, S. J., D. L. Freidenberg, G. W. Paulson, E. C. Shuttleworth and
J. A. Christy (1990). The pattern of depressive symptoms varies with
progression of Parkinson's disease. Journal of Neurology, Neurosurgery
and Psychiatry 53, 275-278.
Huber, S. J. and G. W. Paulson (1987). Memory impairment associated with
progression of Huntington's disease. Cortex 23, 275-283.
Huber, S. J., E. C. Shuttleworth and G. W. Paulson (1986). Dementia in
Parkinson's disease. Archives of Neurology 43, 987-990.
Hunt, G. E., G. F. S. Johnson and I. D. Caterson (1989). The effect of age
on cortisol and plasma dexamethasone concentrations in depressed
patients and controls. Journal of Affective Disorders 17, 21-32.
Huppert, F. A. (1987). Cognitive function. In B. D. Cox (Ed.), Health and
Lifestyle Survey, pp. 43-50. London: Health Promotion Research Trust.
Hurwitz, T. A., C. Clark, E. Murphy, H. Klonoff, W. R. W. Martin and
B. D. Pate (1990). Regional cerebral glucose metabolism in major
depressive disorder. Canadian Journal of Psychiatry 35, 684-688.
Husain, M. M., W. M. McDonald, P. M. Doraiswamy, G. S. Figiel, C. Na,
P. R. Escalona, O. B. Boyko, C. B. Nemeroff and K. R. Krishnan
(1991). A magnetic resonance imaging study of putamen nuclei in major
depression. Psychiatric Research and Neuroimaging 40, 95-99.
Jackson, S. W. (1986). Melancholia and Depression: From Hippocratic Times
to Modern Times. New Haven and London: Yale University Press.
Jacoby, R. J. and R. Levy (1980). Computed tomography in the elderly: 3.
Affective disorders. British Journal of Psychiatry 136, 270-275.
Jacoby, R. J., R. Levy and J. M. Bird (1981). Computed tomography and
the outcome of affective disorder. A follow up study of elderly patients.
British Journal of Psychiatry 139, 288-292.
Jankovic, J. (1987). Pathophysiology and clinical assessment of motor
symptoms in Parkinson's disease. In W. C. Roller (Ed.), Handbook of
Parkinson's Disease, pp. 99-126. New York: Marcel Dekker.
Jaspers, K. (1913). Allgemeine psychopathologie. Berlin: Springer.
Joel, D. and I. Weiner (1994). The organization of the basal
ganglia-thalamocortical circuits: open interconnected rather than closed
segregated. Neuroscience 63, 363-379.
Johnson, G. F., G. Hunt, K. Kerr and I. Caterson (1984). Dexamethasone
References 307
suppression test (DST) and plasma dexamethasone levels in depressed
patients. Psychiatry Research 13, 305-313.
Johnson, J., E. Horwath and M. Weissman (1991). The validity of major
depression with psychotic features based on a community study.
Archives of General Psychiatry 48, 1075-1081.
Johnstone, E. C , J. F. W. Deakin and P. Lawler (1980). The Northwick
Park electroconvulsive therapy trial. Lancet 2, 1317-1320.
Jorm, A. F., A. S. Henderson, D. W. K. Kay and P. A. Jacomb (1991).
Mortality in relation to dementia, depression and social integration in
an elderly community sample. International Journal of Geriatric
Psychiatry 6, 5-11.
Josiassen, R., L. Curry and E. Mancall (1983). Development of
neuropsychological deficits in Huntington's disease.
Neuropsychologia 40, 791-796.
Junque, C , J. Pujol, P. Vendrell, O. Bruna, M. Jodar, J. C. Ribas, J. C.
Vinas, A. Capdevila and J. L. Marti-Vilalta (1990). Leuko-araiosis on
magnetic resonance imaging and speed of mental processing. Archives of
Neurology 47, 151-156.
Kalil, K. (1978). Patch-like termination of thalamic fibers in the putamen of
the rhesus monkey: an autoradiographic study. Brain Research 140,
333-339.
Kapur, S. and J. Mann (1992). Role of the dopaminergic system in
depression. Biological Psychiatry 32, 1-17.
Karayanidis, F. (1989). Parkinson's disease: a conceptualization of
neuropsychological deficits within an information-processing framework.
Biological Psychology 29, 149-179.
Kasahara, H., M. Tanno, H. Yamada, K. Endoh, M. Kobayashi and
A. Karasawa (1993). MRI study of the brain in aged volunteers: T2
high signal intensity lesions and high cortical function. Japanese
Journal of Geriatrics 30, 892-900.
Katz, M. M., E. Robins, J. Croughan, S. Secunda and A. Swann (1982).
Behavioural measurement and drug response characteristics of unipolar
and bipolar depression. Psychological Medicine 12, 25-36.
Katz, M. M., S. K. Secunda, R. M. A. Hirschfeld and S. H. Koslow (1979).
The NIMH Clinical Research Branch Collaborative Program on the
psychobiology of depression. Archives of General Psychiatry 36,
765-771.
Katz, M. M., S. Wetzler, S. H. Koslow and S. K. Secunda (1989). Video
methodology in the study of psychopathology and treatment of
depression. Psychiatric Annals 19, 372-381.
Kawashimi, K. (1979). Clinical study of depressive illness in old age. In
H. Orimo (Ed.), Recent Advances in Gerontology. International
Congress Series No. ^69, pp. 179-180. Amsterdam: Excerpta Medica.
Kay, D. W. K. (1962). Outcome and cause of death in mental disorders of
old age: a long-term follow-up of functional and organic psychosis. Ada
Psychiatrica Scandinavia 38, 249—276.
Kay, D. W. K. and K. Bergmann (1966). A follow-up of a random sample of
elderly people seen at home. Journal of Psychosomatic Research 13,
3-12.
Kellner, C. H., D. R. Rubinow and R. M. Post (1986). Cerebral ventricular
size and cognitive impairment in depression. Journal of Affective
308 References

Disorders 10, 215-219.


Kendell, R. (1983). Diagnosis and classification. In R. Kendell and
A. Zealley (Eds.), Companion to Psychiatric Studies. Edinburgh:
Churchill Livingstone.
Kendell, R. E. (1968). The Classification of Depressive Illness. London:
Oxford University Press.
Kendell, R. E. (1976). The classification of depressions: a review of
contemporary confusion. British Journal of Psychiatry 129, 15-28.
Kendell, R. E. (1981). The present status of ECT. British Journal of
Psychiatry 139, 265-283.
Kendell, R. E. (1989). Clinical validity. In L. N. Robins and J. E. Barrett
(Eds.), The Validity of Psychiatric Diagnosis, pp. 305-323. New York:
Raven Press.
Kessler, R. C, K. A. McGonagle, S. Zhao, C. B. Nelson, M. Hughes,
S. Eshlemann, U. Wittchen and K. Kendler (1994). Lifetime and
12-month prevalence of DSM-III-R psychiatric disorders in the United
States. Archives of General Psychiatry 51, 8-19.
Kiloh, L. G. (1961). Pseudodementia. Acta Psychiatrica Scandinavica 37,
336-351.
Kiloh, L. G., G. Andrews and M. Neilson (1988). The long-term outcome of
depressive illness. British Journal of Psychiatry 153, 752-757.
Kiloh, L. G., G. Andrews, M. Neilson and G. N. Bianchi (1971). The
relationship of the syndromes called endogenous and neurotic
depression. British Journal of Psychiatry 121, 183-196.
Kiloh, L. G., J. R. B. Ball and R. F. Garside (1962). Prognostic factors in
treatment of depressive states with imipramine. British Medical
Journal 1, 1225-1227.
Kiloh, L. G. and R. F. Garside (1963). The independence of neurotic
depression and endogenous depression. British Journal of
Psychiatry 109, 451-463.
Kiloh, L. G. and R. F. Garside (1977). Depression: a multivariate study of
Sir Aubrey Lewis's data on melancholia. Australian and New Zealand
Journal of Psychiatry 11, 149-156.
Kirkpatrick, J. B. and L. A. Hayman (1987). White matter lesions in MR
imaging of clinically healthy brains of elderly subjects: possible
pathologic basis. Radiology 162, 509-511.
Klein, D. E. (1974). Endogenomorphic depression. Archives of General
Psychiatry 31, 447-454.
Klein, D. F. and J. M. Davis (1969). Diagnosis and Drug Treatment of
Psychiatric Disorders. Baltimore: Williams and Wilkins.
Klein, H. E., W. Bender, H. Mayr, A. Niederschweiberer and M. Schmauss
(1984). The DST and its relationship to psychiatric diagnosis, symptoms
and treatment outcome. British Journal of Psychiatry 145, 591-599.
Klerman, G. L. (1971). Clinical research in depression. Archives of General
Psychiatry 24, 305-319.
Klerman, G. L. (1983). Problems in the definition and diagnosis of
depression in the elderly. In L. O. Breslau and M. R. Haig (Eds.),
Depression and Aging: Causes, Care and Consequences, pp. 319-325.
New York: Springer.
Kosugi, Y., M. Sase, H. Kuwatani, N. Kinoshita, T. Momose, J. Nishikawa
and T. Watanabe (1993). Neural network mapping for nonlinear
References 309

stereotactic normalization of brain MR images. Journal of Computer


Assisted Tomography 3, 455-460.
Kovacs, M., A. J. Rush, A. T. Beck and S. D. Hollon (1981). Depressed
outpatients treated with cognitive therapy or pharmacotherapy. A
one-year follow-up. Archives of General Psychiatry 38, 33-39.
Kraepelin, E. (1907). Clinical Psychiatry. New York: Delman.
Kraepelin, E. (1921). Manic-depressive Insanity and Paranoia. Edinburgh:
Livingstone.
Krishnan, K. R. (1993a). Neuroanatomic substrates of depression in the
elderly. Journal of Geriatric Psychiatry and Neurology 6, 39-58.
Krishnan, K. R. R. (1991). Organic bases of depression in the elderly
(review). Annual Review of Medicine 42, 261-266.
Krishnan, K. R. R. (1993b). Neuroanatomic substrates of depression in the
elderly. Journal of Geriatric Psychiatry and Neurology 1, 39-58.
Krishnan, K. R. R., V. Goli, E. H. Ellinwood, R. D. France, D. G. Blazer
and C. B. Nemeroff (1988). Leukoencephalopathy in patients diagnosed
as major depressive. Biological Psychiatry 23, 519-522.
Krishnan, K. R. R., M. M. Husain, W. M. McDonald, P. M. Doraiswamy,
G. S. Figiel, O. B. Boyko, E. H. Ellinwood and C. B. Nemeroff (1990).
In vivo stereological assessment of caudate volume in man: effect of
normal aging. Life Sciences 47, 1325-1330.
Krishnan, K. R. R., W. M. McDonald, P. M. Doraiswamy, L. A. Tupler,
M. Husain, O. B. Boyko, G. S. Figiel and E. H. Ellinwood (1993).
Neuroanatomical substrates of depression in the elderly. European
Archives of Psychiatry and Clinical Neuroscience 243, 41-46.
Krishnan, K. R. R., W. M. McDonald, P. R. Localona, P. M. Doraiswamy,
C. Na, M. M. Husain, G. S. Figiel, O. B. Boyko, E. H. Ellinwood and
C. B. Nemeroff (1992). Magnetic resonance imaging of the caudate
nuclei in depression: preliminary observations. Archives of General
Psychiatry 49, 553-557.
Krishnan, K. R. R., J. C. Ritchie, L. A. Tupler, W. M. McDonald,
D. Knight, C. B. Nancroff and S. Marcovinch (1994). Apolipoprotein
E-e4 - Reply (correspondence). Neurology 44, 2420-2421.
Kuhs, H. and D. Reschke (1992). Psychomotor activity in unipolar and
bipolar depressive patients. Psychopathology 25, 109-116.
Kumar, A., A. Newberg, A. Alavi, J. Berlin, R. Smith and M. Reivich
(1993). Regional cerebral glucose metabolism in late-life depression and
Alzheimer disease: a preliminary positron emission tomography study.
Proceedings of the National Academy of Science 90, 7019-7023.
Kunzle, H. (1975). Bilateral projections from precentral motor cortex to the
putamen and other parts of the basal ganglia. An autoradiographic
study in Macaca fascicularis. Brain Research 88, 1995-2009.
Kupfer, D. J., T. P. Detre, F. G. Foster, G. J. Tucker and J. Delgado (1972).
The application of Delgado's telemetric recorder for human studies.
Behavioural Biology 7, 585-590.
Kupfer, D. J., B. L. Weiss, F. G. Foster, T. P. Detre and R. McPartland
(1974). Psychomotor activity in affective states. Archives of General
Psychiatry 30, 765-768.
Laitinen, L. (1969). Desipramine in the treatment of Parkinson's disease.
Ada Neurologica Scandinavica 45, 109-113.
Lee, A. S. and R. M. Murray (1988). The long-term outcome of Maudsley
310 References

depressives. British Journal of Psychiary 153, 741-751.


Lees, A. J. and E. Smith (1983). Cognitive deficits in the early stages of
Parkinson's disease. Brain 106, 257-270.
Leonhard, K. (1979). The Classification of Endogenous Psychoses (5th ed.),
R. Berman (trans.); E. Robins (Ed.). New York: Irvington Publishers.
Lesser, I. M., I. Mena, K. B. Boone, B. L. Miller, C. M. Mehringer and
M. Wohl (1994). Reduction of cerebral blood flow in older depressed
patients. Archives of General Psychiatry 51, 677-686.
Lesser, I. M., B. L. Miller and K. B. Boone (1991). Brain injury and
cognitive function in late-onset psychotic depression. Journal of
Neuropsychiatric and Clinical Neuroscience 3, 33-40.
Lewis, A. (1934). Melancholia: a clinical survey of depressive states. Journal
of Mental Science 80, 277-378.
Lewis, A. (1938). States of depression: their clinical and aetiological
differentiation. British Medical Journal 2, 875-878.
Lewis, A. (1967). Enquiries in Psychiatry: Clinical and Social Investigations.
London: Rout ledge and Kegan Paul.
Lewis, A. (1971). 'Endogenous' and 'exogenous': a useful dichotomy?
Psychological Medicine 1, 191-196.
Lewis, A. J. (1931). A Clinical and Historical Survey of Depressive States
Based on the Study of 61 Cases. Ph.D. thesis, University of Adelaide,
Adelaide.
Lishman, W. A. (1980). Organic Psychiatry: The Psychological Consequences
of Cerebral Disorder. Oxford: Blackwell Scientific Publications.
Loreck, D. J. and M. F. Folstein (1993). Depression in Alzheimer's disease.
In S. E. Starkstein and R. G. Robinson (Eds.), Depression in Neurologic
Disease, pp. 50-62. Baltimore: Johns Hopkins University Press.
Maas, J., S. Koslow, J. Davis, M. Katz, J. Mendels, E. Robins, P. Stokes and
C. Bowden (1980). Biological components of the NIMH Clinical
Research Branch Collaborative Program on the Psychobiology of
Depression: background and theoretical considerations. Psychological
Medicine 10, 759-776.
Macdonald, A. J. D. (1992). Old age depression and organic brain change.
In T. Arie (Ed.), Recent Advances in Psychogeriatrics 2, pp. 45-58.
Edinburgh: Churchill Livingstone.
Maes, M., E. Bosmans, E. Suy, C. Vandervorst, C. Dejonckheere and
M. Raus (1991). Antiphospholipid, antinuclear, Epstein-Barr and
cytomegalovirus antibodies and soluble interleukin-2 receptors in
depressive patients. Journal of Affective Disorders 21, 133-140.
Maes, M., J. Calabrese and H. Y. Meltzer (1994). The relevance of the in-
versus outpatient status for studies on HPA-axis in depression:
spontaneous hypercortisolism is a feature of major depressed inpatients
and not of major depression per se. Progress in
Neuro-Psychopharmacology and Biological Psychiatry 18, 503—517.
Maes, M., L. Maes, C. Schotte, M. Vandewoude, M. Martin, P. D'Hondt,
P. Blockx, S. Sharpe and P. Cosyns (1990). Clinical subtypes of
unipolar depression: Part III. Quantitative differences in various
biological markers between the cluster-analytically generated nonvital
and vital depression classes. Psychiatry Research 34, 59-75.
Maes, M., H. Meltzer, J. Jacobs, E. Suy, J. Calabrese, B. Minner and
J. Raus (1993). Autoimmunity in depression: increased antiphospholipid
References 311
autoantibodies (review). Ada Psychiatrica Scandanavica 87, 160-166.
Maguire, K. P., I. Schweitzer, N. Biddle, S. Bridge and J. W. G. Tiller
(1987). The dexamethasone suppression test: importance of
dexamethasone concentrations. Biological Psychiatry 22, 957-967.
Maguire, K. P., V. M. Tuckwell, I. Schweitzer, J. W. G. Tiller and B. M.
Davies (1990). Dexamethasone kinetics in depressed patients before and
after clinical response. Psychoneuroendocrinology 15, 113-123.
Mapother, E. (1926). Manic-depressive psychosis. British Medical Journal 2,
872-877.
Marsden, C. D. (1989). Slowness of movement in Parkinson's disease.
Movement Disorders 4, 526-537.
Marshall, V. G., W. G. Bradley, C. E. Marshall, T. Bhoopat and R. H.
Rhodes (1988). Deep white matter infarction: correlation of MR
imaging and histopathologic findings. Radiology 167, 517-522.
Martinot, J., P. Hardy, A. Feline, J.-D. Huret, B. Mazoyer, D. Attar-Levy,
S. Pappata and A. Syrota (1990). Left prefrontal glucose
hypometabolism in the depressed state: a confirmation. American
Journal of Psychiatry 147, 1313-1317.
Mathew, R. J., W. H. Wilson and D. G. Daniel (1985). The effect of
nonsedating doses of diazepam on regional cerebral blood flow.
Biological Psychiatry 20, 1109-1116.
Matison, R., R. Mayeux, J. Rosen and S. Fahn (1982). "Tip-of-the-tongue"
phenomenon in Parkinson's disease. Neurology 32, 567-570.
Matsubayashi, K., K. Shimada, A. Kawamoto and T. Ozawa (1992).
Incidental brain lesions on magnetic resonance imaging and
neurobehavioral functions in apparently healthy elderly. Stroke 23,
175-180.
Maudsley, H. (1895). The Pathology of the Mind: a Study of Its Distempers,
Deformities and Disorders. London and New York: Macmillan.
Maxwell, A. E. (1971). Multivariate statistical methods and classification
problems. British Journal of Psychiatry 119, 121-127.
Mayberg, H. S. (1993). Neuroimaging studies of depression in neurologic
disease. In S. E. Starkstein and R. G. Robinson (Eds.), Depression in
Neurologic Disease, pp. 186-216. Baltimore: Johns Hopkins University
Press.
Mayberg, H. S., P. J. Lewis, W. Regenold and H. N. Wagner (1994).
Paralimbic hypoperfusion in unipolar depression. Journal of Nuclear
Medicine 35, 929-934.
Mayberg, H. S., S. E. Starkstein, C. E. Peyser, J. Brandt, R. F. Dannals and
S. E. Folstein (1992). Paralimbic frontal lobe hypometabolism in
depression associated with Huntington's disease. Neurology 42,
1791-1797.
Mayberg, H. S., S. E. Starkstein, B. Sadzot, T. Preziosi, P. L. Andrezejewski,
R. F. Dannals, H. N. J. Wagner and R. G. Robinson (1990b). Selective
hypometabolism in the inferior frontal lobe in depressed patients with
Parkinson's disease. Annals of Neurology 28, 57-64.
Mayeux, R. (1990). Depression in the patient with Parkinson's disease.
Journal of Clinical Psychiatry 51 (Suppl. 7), 20-23.
Mayeux, R., Y. Sten, J. Rosen and J. Leventhal (1981). Depression,
intellectual impairment and Parkinson's disease. Neurology 31, 645-650.
McDonald, R. (1985). Factor Analysis and Related Methods. Hillsdale:
312 References
Lawrence Erlbaum.
McLachlan, G. J. (1992). The use of cluster analysis and related techniques
in medicine. Statistical Methods in Medical Research 1, 27-48.
McLachlan, G. J. and K. E. Basford (1988). Mixture Models: Inference and
Applications to Clustering. New York: Marcel Dekker.
Meador-Woodruff, J. H., J. F. Greden, L. Grunhaus and R. F. Haskett
(1990). Severity of depression and hypothalamic-pituitary-adrenal axis
dysregulation: identification of contributing factors. Ada Psychiatrica
Scandinavica 81, 364-371.
Meehl, P. E. (1992). Factors and taxa, traits and types, differences of degree
and differences of kind. Journal of Personality 60, 117-174.
Mendels, J. and C. Cochrane (1968). The nosology of depression: the
endogenous-reactive concept. American Journal of Psychiatry 124,
1-11 supplement.
Merikangas, K. R., N. J. Risch and M. M. Weissman (1994). Comorbidity
and co-transmission of alcoholism, anxiety and depression. Psychological
Medicine 24, 69-80.
Merskey, H. (1991). Shell-shock. In G. E. Berrios and H. Freeman (Eds.), 150
Years of British Psychiatry, 1841-1991, pp. 245-267. London: Gaskell.
Meyers, B. S. and R. Greenberg (1986). Late-life delusional depression.
Journal of Affective Disorders 11, 133-137.
Meyers, B. S., R. Greenberg and V. Mei-Tal (1985). Delusional depression in
the elderly. In C. A. Shamoian (Ed.), Treatment of Affective Disorders
in the Elderly, pp. 19-28. Washington, DC: American Psychiatric Press.
Miller, K. B. and J. C. Nelson (1987). Does the dexamethasone suppression
test relate to subtypes, factors, symptoms, or severity? Archives of
General Psychiatry 44, 769-774.
Miller, W. R. (1975). Psychological deficits in depression. Psychological
Bulletin 82, 238-260.
Minter, R. E. and M. R. Mandel (1979). The treatment of psychotic major
depressive disorder with drugs and electroconvulsive therapy. Journal of
Nervous and Mental Disease 167, 726-733.
Mitchell, P., G. Parker, K. Jamieson, K. Wilhelm, I. Hickie, H. Brodaty,
P. Boyce, D. Hadzi-Pavlovic and K. Roy (1992). Are there any
differences between bipolar and unipolar melancholia? Journal of
Affective Disorders 25, 97-106.
Montgomery, S. and M. Asberg (1979). A new depression scale designed to
be sensitive to change. British Journal of Psychiatry 134, 382-389.
Moran, P. A. P. (1966). The establishment of a psychiatric syndrome.
British Journal of Psychiatry 112, 1165-1171.
Morris, R. G., J. J. Downes, B. J. Sahakian, J. L. Levenden, A. Heald and
T. W. Robbins (1988). Planning and spatial working memory in
Parkinson's disease. Journal of Neurology, Neurosurgery and
Psychiatry 51, 757-766.
Mortimer, J. A., F. J. Pirozzolo, E. C. Hansch and D. D. Webster (1982).
Relationship of motor symptoms to intellectual deficits in Parkinson
disease. Neurology 32, 133-137.
Moss, M. B., M. S. Albert, N. Butters and M. Payne (1986). Differential
patterns of memory loss among patients with Alzheimer's disease,
Huntington's disease and alcoholic Korsakoff's syndrome. Archives of
Neurology 43, 239-246.
References 313
Mullaney, J. A. (1984). The relationship between anxiety and depression: a
review of some principal component analytic studies. Journal of
Affective Disorders 7, 139-148.
Murphy, D. G. M., C. DeCarli, M. B. Schapiro, S. I. Rapoport and
B. Horwitz (1992). Age-related differences in volumes of subcortical
nuclei, brain matter, and cerebrospinal fluid in healthy men as measured
with magnetic resonance imaging. Archives of Neurology 49, 839-845.
Murphy, D. L., G. M. Henry and H. Weingartner (1972). Catecholamines
and memory: enhanced verbal learning during L-DOPA administration.
Psychopharmacologia 27, 319-326.
Murphy, E. (1983). The prognosis of depression in old age. British Journal
of Psychiatry 142, 111-119.
Murphy, E., R. Smith, J. Lindesay and J. Slattery (1988). Increased
mortality rates in late life depression. British Journal of
Psychiatry 152, 347-353.
Murphy, E. A. (1964). One cause? Many causes? The argument from the
bimodal distribution. Journal of Chronic Disorders 17, 301-324.
Murphy, J. M., D. M. Berwick and M. C. Weinstein (1987). Performance of
screening and diagnostic tests: application of receiver operating
characteristic analysis. Archives of General Psychiatry 44, 550-555.
Musetti, L., G. Perugi, A. Soriani, V. M. Rossi, G. B. Cassano and H. S.
Akiskal (1989). Depression before and after age 65. British Journal of
Psychiatry 155, 330-336.
Nelson, J. C. and D. S. Charney (1981). The symptoms of major depressive
illness. American Journal of Psychiatry 138, 1-13.
Nelson, J. C , Y. Conwell, K. Kim and C M . Mazure (1989). Age at onset in
late-life delusional depression. American Journal of Psychiatry 146,
785-786.
Nelson, W. H., A. Khan and W. W. Orr Jr. (1984). Delusional depression:
phenomenology, neuroendocrine function, and tricyclic antidepressant
response. Journal of Affective Disorders 6, 297-306.
Newman, R. P., H. Weingartner, S. A. Smallberg and D. B. Calne (1984).
Effortful and automatic memory: effects of dopamine. Neurology 34,
805-807.
Newton, R. D. (1948). The identity of Alzheimer's disease and senile
dementia and their relationship to senility. Journal of Mental
Science 94, 225-249.
Ni Brolchain, M. (1979). Psychotic and neurotic depression: I. Some points
of method. British Journal of Psychiatry 134, 87-93.
Nystrom, S. (1965). On the relation between clinical factors and efficacy of
ECT in depression. Ada Psychiatrica Scandinavia 181, 121-134.
O'Brien, J. T. and D. Ames (1994). Editorial: Why do the depressed elderly
die? International Journal of Geriatric Psychiatry 9, 689-693.
Parker, G. and M.-P. Austin (1995). A clinical perspective on SPECT.
Australian and New Zealand Journal of Psychiatry 29, 38-47.
Parker, G., P. Boyce, D. Mitchell, D. Hadzi-Pavlovic, K. Wilhelm, I. Hickie
and H. Brodaty (1992a). Comparison of clinician rated and family
corroborative witness data for depressed patients. Journal of Affective
Disorders 24, 25-34.
Parker, G. and D. Hadzi-Pavlovic (1993). Old data, new interpretation: a
re-analysis of Sir Aubrey Lewis' MD Thesis. Psychological Medicine 23,
314 References
859-870.
Parker, G., D. Hadzi-Pavlovic, M.-P. Austin, P. Mitchell, K. Wilhelm,
I. Hickie, P. Boyce and K. Eyers (1995a). Sub-typing depression: I. Is
psychomotor disturbance necessary and sufficient to define melancholia.
Psychological Medicine (in press).
Parker, G., D. Hadzi-Pavlovic and P. Boyce (1989). Endogenous depression
as a construct: a quantitative analysis of the literature and a study of
clinician judgements. Australian and New Zealand Journal of
Psychiatry 23, 357-368.
Parker, G., D. Hadzi-Pavlovic, P. Boyce, K. Wilhelm, H. Brodaty,
P. Mitchell, I. Hickie and K. Eyers (1990). Classifying depression by
mental state signs. British Journal of Psychiatry 157, 55-65.
Parker, G., D. Hadzi-Pavlovic, H. Brodaty, M.-P. Austin, P. Mitchell,
K. Wilhelm and I. Hickie (1995b). Sub-typing depression: II. Clinical
distinction of psychotic depression and non-psychotic melancholia.
Psychological Medicine (in press).
Parker, G., D. Hadzi-Pavlovic, H. Brodaty, P. Boyce, P. Mitchell,
K. Wilhelm and I. Hickie (1992b). Predicting the course of melancholic
and non-melancholic depression: a naturalistic comparison study.
Journal of Nervous and Mental Disease 180, 693-702.
Parker, G., D. Hadzi-Pavlovic, H. Brodaty, P. Boyce, P. Mitchell,
K. Wilhelm, I. Hickie and K. Eyers (1993). Psychomotor disturbance in
depression: defining the constructs. Journal of Affective Disorders 27,
255-265.
Parker, G., D. Hadzi-Pavlovic, I. Hickie, P. Boyce, P. Mitchell, K. Wilhelm
and H. Brodaty (1991a). Distinguishing psychotic and non-psychotic
melancholia. Journal of Affective Disorders 22, 135-148.
Parker, G., D. Hadzi-Pavlovic, I. Hickie, H. Brodaty, P. Boyce, P. Mitchell
and K. Wilhelm (1995c). Sub-typing depression: III. Development of a
clinical algorithm for melancholia and comparison with other diagnostic
measures. Psychological Medicine (in press).
Parker, G., D. Hadzi-Pavlovic, I. Hickie, P. Mitchell, K. Wilhelm,
H. Brodaty, P. Boyce, K. Eyers and F. Pedic (1991b). Psychotic
depression: a review and clinical practice. Australian and New Zealand
Journal of Psychiatry 25, 63-76.
Parker, G., D. Hadzi-Pavlovic, P. Mitchell, I. Hickie, K. Wilhelm,
H. Brodaty, P. Boyce and K. Roy (1991d). Psychosocial risk factors
distinguishing melancholic and nonmelancholic depression: a comparison
of six systems. Psychiatry Research 39, 211-226.
Parker, G., D. Hadzi-Pavlovic, K. Wilhelm, I. Hickie, H. Brodaty, P. Boyce,
P. Mitchell and K. Eyers (1994). Defining melancholia: properties of a
refined sign-based measure. British Journal of Psychiatry 164, 316—326.
Parker, G., W. Hall, P. Boyce, D. Hadzi-Pavlovic, P. Mitchell, K. Wilhelm,
H. Brodaty, I. Hickie and K. Eyers (1991c). Depression sub-typing:
unitary, binary or arbitrary? Australian and New Zealand Journal of
Psychiatry 25, 63-76.
Parker, G., K. Roy, D. Hadzi-Pavlovic and F. Pedic (1992c). Psychotic
(delusional) depression: a meta-analysis of physical treatments. Journal
of Affective Disorders 24, 17-24.
Parkes, C. M. (1973). Bereavement. London: Tavistock Publications.
Parkinson, J. (1817). An Essay on the Shaking Palsy. London: Sherwood,
References 315
Neely & Jones.
Paykel, E. S. (1971). Classification of depressed patients: a cluster analysis
derived grouping. British Journal of Psychiatry 118, 275-288.
Paykel, E. S., B. Prusoff and G. L. Klerman (1971). The
endogenous-neurotic continuum in depression: rater independence and
factor distributions. Journal of Psychiatric Research 8, 73-90.
Pearlson, G. D. and A. E. Veroff (1981). Computerised tomographic scan
changes in manic-depressive illness [letter]. Lancet 2, 470.
Pearson, K. (1894). Contributions to the mathematical theory of evolution.
Philosophy Transactions A 185, 71-110.
Pendleton Jones, B., M. Henderson and C. A. Welch (1987). Executive
functions in unipolar depression before and after electroconvulsive
therapy. International Journal of Neuroscience 38, 287-297.
Perris, C. (1966). A study of bipolar (manic-depressive) and unipolar
recurrent depressive psychoses. Ada Psychiatrica Scandinavica 42
(Suppl. 194), 1-189.
Perris, C. (1990). The importance of Karl Leonhard's classification of
endogenous psychoses. Psychopathology 23, 282-290.
Perris, C. (1992). The distinction between unipolar and bipolar mood
disorders. L'Encephale XVIII, 9-13.
Perry, E. K. (1986). The cholinergic hypothesis - ten years on. British
Medical Bulletin 42, 63-69.
Peyser, C. E. and S. E. Folstein (1993). Depression in Huntington's disease.
In S. E. Starkstein and R. G. Robinson (Eds.), Depression in Neurologic
Disease, pp. 117-138. Baltimore: Johns Hopkins University Press.
Philipp, M., W. Maier and F. Holsboer (1986). Psychopathological correlates
of plasma cortisol after dexamethasone suppression: a polydiagnostic
approach. Psychoneuroendocrinology 11, 499-507.
Pichot, P., C. B. Pull, E. Nowicka, M. C. Pull and R. Von Frenckell (1979).
Apropos de la melancholie d'involution: IV. — comparison de la
symptomatologie des depressions unipolaires a debut precoce et de
depressions unipolaires a debut tardif [Involutional melancholia: IV. —
comparison of the symptomatology of unipolar depressions with early
onset and unipolar depressions with late onset]. Annales
Medico-Psychologiques 137, 968-986.
Pillon, B., B. Dubois, F. Lhermitte and Y. Agid (1986). Heterogeneity of
cognitive impairment in progressive supranuclear palsy. Parkinson's
disease, and Alzheimer's disease. Neurology 36, 1179-1185.
Pilowsky, I. and M. Katsikitis (1994). The classification of facial emotions: a
computer-based taxonomic approach. Journal of Affective Disorders 30,
61-71.
Pilowsky, L. S., D. C. Costa, P. J. Ell, R. M. Murray, N. P. L. G. Verhoeff
and R. W. Kerwin (1992). Clozapine, single photon emission
tomography, and the d2 dopamine receptor blockage hypothesis of
schizophrenia. Lancet 340, 199-202.
Pirozzolo, F. J., E. C. Hansch, J. A. Mortimer, D. D. Webster and M. A.
Kuschowski (1982). Dementia in Parkinson's disease: a
neuropsychological analysis. Brain Cognition 1, 71-83.
Poirier, J., J. Davignon, D. Bouthillier, S. Kogan, P. Bertrand and
S. Gauthiers (1993). Apolipoprotein E polymorphism and Alzheimer's
disease. Lancet 342, 697-699.
316 References

Popescu, C , R. Ionescu, I. Jipescu and S. Popa (1991). Psychomotor


functioning in unipolar and bipolar affective disorders. Romanian
Journal of Neurology and Psychiatry 29, 17-33.
Post, F. (1962). The Significance of Affective Disorders in Old Age.
Maudsley Monographs, No. 10. London: Oxford University Press.
Post, F. (1972). The management and nature of depressive illness in late life:
a follow-through study. British Journal of Psychiatry 121, 393-404.
Rabins, P. V., G. D. Pearlson and E. Aylward (1991). Cortical magnetic
resonance imaging changes in elderly patients with major depression.
American Journal of Psychiatry 148, 617-620.
Radloff, L. S. (1977). The CES-D scale: a self-report depression scale for
research in the general population. Applied Psychological
Measurement 1, 385-401.
Rafal, R. D., M. I. Posner, J. A. Walker and F. J. Friederich (1984).
Cognition and the basal ganglia: separating mental and motor
components of performance in Parkinson's disease. Brain 107,
1083-1094.
Rao, S. M., G. J. Leo and P. St. Aubin-Faubert (1989). On the nature of
memory disturbance in multiple sclerosis. Journal of Clinical and
Experimental Neuropsychology 11, 699-712.
Raskin, A., A. S. Friedman and A. Di Mascio (1982). Cognitive and
performance deficits in depression. Psychopharmacology Bulletin 18,
196-206.
Raskin, A., J. Schulterbrandt, N. Reatig and J. J. McKeon (1969).
Replication of factors of psychopathology in interview, ward behavior
and self-report ratings of hospitalized depressives. The Journal of
Nervous and Mental Disease 148, 87-98.
Raskin, A., J. Schulterbrandt, N. Reatig and C. E. Rice (1967). Factors of
psychopathology in interview, ward behavior, and self-report ratings of
hospitalized depressives. Journal of Consulting Psychology 31, 270-278.
Regier, D. A., M. E. Farmer, D. S. Rae, J. K. Myers, M. Kramer, L. N.
Robins, L. K. George, M. Karno and R. S. Lock (1993). One-month
prevalence of mental disorders in the United States and socio-
demographic characteristics: the Epidemiologic Catchment Area study.
Acta Psychiatrica Scandinavica 88, 35-47.
Regier, D. A., J. K. Myers, M. Kramer, L. N. Robins, D. G. Blazer, R. L.
Hough, W. W. Eaton and B. Z. Locke (1984). The NIMH Epidemiologic
Catchment Area program. Archives of General Psychiatry 41, 934-941.
Rehm, L. P. (1987). The measurement of behavioral aspects of depression.
In A. J. Marsella, R. M. A. Hirschfeld, and M. M. Katz (Eds.), The
Measurement of Depression, pp. 199-239. New York: Guilford Press.
Reitan, R. M. (1958). Validity of the trail-making test as an indicator of
organic brain damage. Perceptual and Motor Skills 8, 271-276.
Reus, V. I., E. Silberman, R. M. Post and H. Weingartner (1979).
D-amphetamine: effects on memory in a depressed population.
Biological Psychiatry 14, 345-356.
Robbins, T. W., E. M. Joyce and B. J. Sahakian (1992). Neuropsychology
and imaging. In E. S. Paykel (Ed.), Handbook of Affective Disorders
(2nd ed.), pp. 289-309. New York: Guilford Press.
Robins, L. N., J. E. Helzer, J. Croughan and K. Ratcliff (1981). National
Institute of Mental Health Diagnostic Interview Schedule. Archives of
References 317

General Psychiatry 38, 381-389.


Robins, L. N., J. E. Helzer, M. M. Weissman, H. Orvaschel, E. Gruenberg,
J. D. Burke and D. A. Regier (1984). Lifetime prevalence of specific
psychiatric disorders in three sites. Archives of General Psychiatry 41,
949-958.
Robins, L. N., J. Wing, H. U. Wittchen, J. E. Helzer, T. F. Babor, J. Burke,
A. Farmer, A. Jablenski, R. Pickens, D. A. Regier, N. Sartorius and
L. H. Towle (1988). The Composite International Diagnostic Interview:
an epidemiological instrument suitable for use in conjunction with
different diagnostic systems and in different cultures. Archives of
General Psychiatry 45, 1069-1077.
Robinson, R. G., K. L. Kubos, L. B. Starr, K. Rao and T. R. Price (1984).
Mood disorders in stroke patients: importance of location of lesion.
Brain 107, 81-93.
Robinson, R. G., R. M. Parikh and R. A. Rodriguez (1988a). CSF
monoamine metabolite measures in post-stroke depression. Social
Neuroscience Abstracts S7, 301.
Robinson, R. G., S. E. Starkstein and T. R. Price (1988b). Post stroke
depression and lesion location. Stroke 19, 125-126.
Roeder, K. (1994). A graphical technique for determining the number of
components in a mixture of normals. Journal of the American Statistical
Association 89, 487-495.
Rogers, D. (1985). The motor disorders of severe psychiatric illness. A
conflict of paradigms. British Journal of Psychiatry 147, 221-232.
Rogers, D. (1992). Bradyphrenia in Parkinson's disease. In S. J. Huber and
J. L. Cummings (Eds.), Parkinson's Disease: Neurobehavioral Aspects,
pp. 86-96. New York: Oxford University Press.
Rogers, D., A. J. Lees, E. Smith, M. Trimble and G. M. Stern (1987).
Bradyphrenia in Parkinson's disease and psychomotor retardation in
depressive illness: an experimental study. Brain 110, 761-776.
Rogers, D., A. J. Lees, M. Trimble and G. M. Stern (1986). Concept of
bradyphrenia: a neuropsychiatric approach. Advances in Neurology 45,
447-450.
Ron, M. A. and S. J. Logsdail (1989). Psychiatric morbidity in multiple
sclerosis: a clinical and MRI study. Psychological Medicine 19, 887-895.
Rosenthal, S. H. and J. E. Gudeman (1967). The endogenous depressive
pattern. An empirical investigation. Archives of General Psychiatry 16,
241-249.
Rosenthal, S. M. and G. L. Klerman (1966). Content and consistency in the
endogenous depressive pattern. British Journal of Psychiatry 112,
471-484.
Ross, T. A. (1924). The Common Neuroses. London: Edward Arnold.
Roth, M. (1955). The natural history of mental disorder in old age. Journal
of Mental Science 110, 281-301.
Roth, M., C. Gurney, R. F. Garside and T. A. Kerr (1972). Studies in the
classification of affective disorders. The relationship between anxiety
states and depressive illnesses. British Journal of Psychiatry 121,
147-161.
Rovner, B. W., J. Broadhead, M. Spencer, K. Carson and M. F. Folstein
(1989). Depression and Alzheimer's disease. American Journal of
Psychiatry 146, 350-353.
318 References

Roy, A., D. Pickar, M. Linnoila and W. Z. Potter (1985). Plasma


norepinephrine level in affective disorders: relationship to melancholia.
Archives of General Psychiatry 42, 1181-1185.
Roy-Byrne, P. P., H. Weingartner, L. M. Bierer, K. Thompson and R. M.
Post (1986). Effortful and automatic cognitive processes in depression.
Archives of General Psychiatry 43, 265-267.
Royant-Parola, S., A. A. Borbely, I. Tobler, O. Benoit and D. Widlocher
(1986). Monitoring of long-term motor activity in depressed patients.
British Journal of Psychiatry 149, 288-293.
Ruegg, R. G., S. Zisook and N. R. Swedlow (1988). Depression in the aged,
an overview. Psychiatric Clinics of North America 11, 83-99.
Rush, A. J. and J. E. Weissenburger (1994). Melancholic symptom features
and DSM-IV. American Journal of Psychiatry 151, 489-498.
Rush, A. J. and J. E. Weissenburger (1995). Melancholic symptom features:
a review and options for DSM-IV. In T. A. Widiger, A. J. Prances,
M. A. Pincus, R. Ross, M. B. First, and W. Davis (Eds.), DSM-IV
Source Book. Volume 2. Washington, DC: American Psychiatric Press.
Sachdev, P. and A. M. Aniss (1994). Slowness of movement in melancholic
depression. Biological Psychiatry 35, 253-262.
Sackeim, H. A., I. Prohovnik, J. R. Moeller, R. P. Brown, S. Apter,
J. Prudic, D. P. Devanand and S. Mukherjee (1990). Regional cerebral
blood flow in mood disorders: I. Comparison of major depressives and
normal controls at rest. Archives of General Psychiatry 47, 60-70.
Sackeim, H. A., I. Prohovnik, J. R. Moeller, R. Mayeux, Y. Stern and D. P.
Devanand (1993). Regional cerebral blood flow in mood disorders: II.
Comparison of major depression and Alzheimer's disease. The Journal
of Nuclear Medicine 34, 1090-1101.
Sands, J. R. and M. Harrow (1994). Psychotic unipolar depression at
follow-up: factors related to psychosis in the affective disorders.
American Journal of Psychiatry 151, 995-1000.
Santamaria, J., E. Tolosa and A. Valles (1986). Parkinson's disease with
depression: a possible subgroup of idiopathic parkinsonism.
Neurology 36, 1130-1133.
Saunders, A. M. and A. D. Roses (1993). Apolipoprotein E4 allele frequency,
ischemic cerebrovascular disease, and Alzheimer's disease, [letter].
Stroke 24, 1416-1417.
Saunders, A. M., W. Strittmatter, D. Schmechel, P. H. St George-Hislopp,
M. A. Pericak-Vance and S. H. Joo (1993). Association of
apolipoprotein E allele e4 with late-onset familial and sporadic
Alzheimer's disease. Neurology 43, 1467-1472.
Schatzberg, M. D. and A. J. Rothschild (1992). Psychotic (delusional) major
depression: should it be included as a distinct syndrome in DSM-IV.
American Journal of Psychiatry 149, 733-745.
Schiffer, R. B., R. Kurlan, A. Rubin and S. Boer (1988). Evidence for
atypical depression in Parkinson's disease. American Journal of
Psychiatry 145, 1020-1022.
Schlegel, S. and K. Kretzschmar (1987). Computed tomography in affective
disorders. Part 1. Ventricular and sulcal measurements. Biological
Psychiatry 22, 4-14.
Schlegel, S., D. Nieber, B. Solbach and P. Doebereiner (1989). Are ratings of
psychomotor retardation correlated with neuropsychological
References 319
measurements? Psychopathology 22, 194-197.
Sergent, J. (1994). Brain-imaging studies of cognitive functions. Trends in
Neuroscience 17, 221-227.
Shallice, T. (1988). From Neuropsychology to Mental Structure. Cambridge:
Cambridge University Press.
Shima, S., T. Shikano, T. Kitamuna, Y. Masuda, T. Tsukumo, S. Kanaba
and M. Asai (1984). Depression and ventricular enlargement. Acta
Psychiatrica Scandinavica 70, 275-277.
Shinfuki, N. (1979). Mental health of the aged (author's translation). Nippon
Ronen Igakkai Zasshi 15, 336-340.
Silberman, E. K., H. Weingartner and R. M. Post (1983). Thinking disorder
in depression. Archives of General Psychiatry 40, 775-780.
Silfverskiold, P. and J. Risberg (1989). Regional cerebral blood flow in
depression and mania. Archives of General Psychiatry 46, 253-259.
Simon, H., B. Scatton and M. Le Moal (1980). Dopaminergic A10 neurons
are involved in cognitive functions. Nature 286, 150-151.
Smith, A. (1968). The symbol-digit modalities test: a neuropsychologic test
of learning and other cerebral disorders. In J. Helmuth (Ed.), Learning
Disorders, pp. 83-91. Seattle: Special Child Publications.
Smith, J., V. Carr, H. Morris and J. Gilliland (1988). The dexamethasone
suppression test in relation to symptomatology: preliminary findings
controlling for serum dexamethasone concentrations. Psychiatry
Research 25, 123-133.
Snaith, R. P., G. Bridge and M. Hamilton (1976). The Leeds scale for the
self-assessment of anxiety and depression. British Journal of
Psychiatry 128, 156-165.
Spiker, D. G., J. C. Weiss, R. S. Dealy, S. J. Griffin, I. Hanin, J. F. Neil,
J. M. Perel, A. J. Rossi and P. H. Soloff (1985). The pharmacological
treatment of delusional depression. American Journal of
Psychiatry 142, 430-435.
Spitzer, R. L., J. Endicott and E. Robins (1980). Research Diagnostic
Criteria (RDC) for a Selected Group of Functional Disorders (3rd ed.).
New York: New York State Psychiatric Institute.
Spitzer, R. L. and J. B. Williams (1983). Development of a Diagnostic
Clinical Interview for Psychiatric Research (Technical Proposal RFP
NIMH DB-83-0007). Rockville, MD: National Institute of Mental
Health.
Spreen, O. and A. L. Benton (1969). Neurosensory Center Comprehensive
Examination for Aphasia. Victoria, BC: Neuropsychological Laboratory
Department of Psychology, University of Victoria.
Staner, L., P. Linkowski and J. Mendlewicz (1994). Biological markers as
classifiers for depression: a multivariate study. Progress in
Neuro-Psychopharmacology and Biological Psychiatry 18, 899-914.
Staner, L., M. Maes, E. Bouillon and P. Linkowski (1992). Biological
correlates of the Newcastle Scale in depressive illness: a multivariate
approach. Acta Psychiatrica Scandinavica 85, 345-350.
Starkstein, S. E., J. Brandt, S. Folstein, M. Strauss, M. L. Berthier, G. D.
Pearlson, D. Wong, A. McDonnell and M. Folstein (1988).
Neuropsychological and neuroradiological correlates in Huntington's
disease. Journal of Neurology; Neurosurgery and Psychiatry 57,
1259-1263.
320 References
Starkstein, S. E. and H. S. Mayberg (1993). Depression in Parkinson's
disease. In S. E. Starkstein and R. G. Robinson (Eds.), Depression in
Neurologic Disease, pp. 97-116. Baltimore: Johns Hopkins University
Press.
Starkstein, S. E., T. J. Preziosi, M. L. Berthier, P. L. Bolduc, H. S. Mayberg
and R. G. Robinson (1989). Depression and cognitive impairment in
Parkinson's disease. Brain 112, 1141-1153.
Starkstein, S. E. and R. G. Robinson (Eds.) (1993). Baltimore: Johns
Hopkins University Press.
Stendstedt, A. (1959). Involutional melancholia: an etiologic, clinical and
social study of endogenous depression in later life, with special reference
to genetic factors. Ada Psychiatrica Scandinavica 34 (Suppl. 127),
1-76.
Sternberg, D. E. and M. E. Jarvik (1976). Memory functions in depression.
Archives of General Psychiatry 33, 219-224.
Stoudemire, A., C. Hill, L. R. Gulley and R. Morris (1989).
Neuropsychological and biomedical assessment of depression-dementia
syndromes. Journal of Neuropsychiatry 1, 347-361.
Strang, R. R. (1965). Imipramine in the treatment of parkinsonism: a
double-blind placebo study. British Medical Journal 2, 33-34.
Strange, P. G. (1993). Dopamine receptors in the basal ganglia: relevance to
Parkinson's disease. Movement Disorders 8, 263-270.
Strauss, J. S. (1994). Is biological psychiatry building on an adequate base?
Clinical reality and underlying processes in schizophrenic disorders. In
N. C. Andreasen (Ed.), Schizophrenia. From Mind to Molecule, pp.
31-44. Washington, DC: American Psychiatric Press.
Strittmatter, W., A. M. Saunders, D. Schmechel, M. A. Pericak-Vance, E. J.
and G. S. Salvesen (1993). Apolipoprotein E: high avidity binding to
beta-amyloid and increased frequency of type 4 allele in late onset
familial Alzheimer's disease. Proceedings of the National Academy of
Science, U.S.A. 90, 1977-1981.
Stromgren, L. S. (1977). The influence of depression on memory. Acta
Psychiatrica Scandinavica 56, 109-128.
Stuss, D. T. and D. F. Benson (1984). Neuropsychological studies of the
frontal lobes. Psychological Bulletin 95, 3-28.
Sze, G., S. J. De Armond, M. Brant-Zawadzki, R. L. Davis, D. Norman and
T. H. Newton (1981). Foci of MRI signal anterior to the frontal horns:
histologic correlation of a normal finding. American Journal of
Neuroradiology 147, 331-337.
Targum, S. D., L. N. Rosen, L. E. DeLisi, D. R. Weinburger and C. M.
Citrin (1983). Cerebral ventricular size in major depressive disorder.
Association with delusional symptoms. Biological Psychiatry 18,
329-336.
Taylor, A. E., J. A. Saint-Cyr and A. E. Lang (1986). Frontal lobe
dysfunction in Parkinson's disease. Brain 109, 845-883.
Teicher, M. H., N. I. Barber, J. M. Lawrence and R. J. Baldessarini (1989).
Motor activity and antidepressant drugs: a proposed approach to
categorizing depression syndromes and their animal models. In G. F.
Koob, C. L. Ehlers, and D. J. Kupfer (Eds.), Animal Models of
Depression. Boston: Birkhauser.
Thase, M., M. Hersen, A. Bellack, J. Himmelhoch and D. Kupfer (1983).
References 321

Validation of a Hamilton subscale for endogenomorphic depression.


Journal of Affective Disorders 5, 267-276.
Torack, R. M. and J. C. Morris (1988). The association of ventral tegmental
area histopathology with adult dementia. Archives of Neurology 45,
497-501.
Upadhyaya, A. K., M. T. Abou-Saleh, K. Wilson, S. J. Grime and
M. Critchley (1990). A study of depression in old age using
single-photon emission computerised tomography. British Journal of
Psychiatry 157 (Suppl. 9), 76-81.
van Praag, H. (1993). Make-believes in psychiatry or the perils of progress.
New York: Brunner/Mazel.
van Praag, H. M., J. Korf, J. P. W. F. Lakke and T. Schut (1975).
Dopamine metabolism in depression, psychoses and Parkinson's disease:
the problem of specificity of biological variables in behaviour disorders.
Psychological Medicine 5, 138-146.
van Praag, H. M., J. Korf and D. Schut (1973). Cerebral monoamines and
depression: an investigation with the probenecid technique. Archives of
General Psychiatry 28, 827-831.
Waldemar, G., S. G. Hasselbalch, A. R. Andersen, F. Delecluse, P. Petersen,
A. Johnsen and O. B. Paulson (1991). " m Tc-d, 1-HMPAO and SPECT
of the brain in normal aging. Journal of Cerebral Blood Flow and
Metabolism 11, 508-521.
Watts, F. N. (1992). Problems of memory and concentration. In C. G.
Costello (Ed.), Symptoms of Depression, pp. 113-140. New York: John
Wiley and Sons.
Webster, D. D. (1968). Critical analysis of the disability in Parkinson's
disease. Modern Treatment 5, 257-282.
Wechsler, D. (1987). Wechsler Memory Scale - Revised. New York: The
Psychological Corporation, Harcourt Brace Jovanovich Inc.
Weckowicz, T. E., K. A. Yonge and A. J. Cropley (1971). Objective therapy
predictors in depression: a multivariate approach. Journal of Clinical
Psychology 27, 4-29.
Weinberger, D. R., K. F. Berman, M. Iadarola, N. Driesen and R. F. Zee
(1988). Prefrontal cortical blood flow and cognitive function in
Huntington's disease. Journal of Neurology, Neurosurgery and
Psychiatry 51, 94-104.
Weiner, M. F., S. D. Edland and H. Luszczynska (1994). Prevalence and
incidence of major depression in Alzheimer's disease. American Journal
of Psychiatry 151, 1006-1009.
Weingartner, H., S. Burns, R. Diebel and P. A. Le Witt (1984). Cognitive
impairments in Parkinson's disease: distinguishing between
effort-demanding and automatic cognitive processes. Psychiatry
Research 11, 223-235.
Weiss, B., D. J. Kupfer, F. G. Foster and J. Delgado (1974). Psychomotor
activity, sleep, and biogenic amine metabolites in depression. Biological
Psychiatry 9, 45-53.
Widlocher, D. J. (1983). Psychomotor retardation: clinical, theoretical, and
psychometric aspects. Psychiatric Clinics of North America 6, 27-40.
Williams, J. B. W. (1988). A structured interview guide for the Hamilton
Rating Scale. Archives of General Psychiatry 45, 742-747.
Willner, P., R. Muscat, M. Papp and D. Sampson (1991). Dopamine,
322 References

depression and anti-depressant drugs. In P. Willner and


J. Scheel-Kruger (Eds.), The Mesolimbic Dopamine System: From
Motivation to Action, pp. 387-410. New York: John Wiley and Sons.
Wilson, R. S., A. W. Kaszniak, H. L. Klawans and D. C. Garron (1980).
High speed memory scanning in parkinsonism. Cortex 12, 67-72.
Wing, J. K., J. E. Cooper and N. Sartorious (1974). Measurement and
Classification of Psychiatric Symptoms: An Instruction Manual for the
PSE and CATE GO Program. London: Cambridge University Press.
Winokur, G., D. Behar and M. Schesser (1980). Clinical and biological
aspects of depression in the elderly. In J. O. Cole and J. E. Barrett
(Eds.), Psychopathology in the Aged, pp. 145-153. New York: Raven
Press.
Winokur, G., W. Coryell, M. Keller, J. Endicott and H. Akiskal (1993). A
prospective follow-up of patients with bipolar and primary unipolar
affective disorder. Archives of General Psychiatry 50, 457-465.
Wittchen, H.-U., G. Knauper and R. C. Kessler (1994). Depression across
the life-cycle. British Journal of Psychiatry — Supplement 26, 16-22.
Wolfe, J., E. Granholm, N. Butters, E. Saunders and D. Janowski (1987).
Verbal memory deficits associated with major affective disorders: a
comparison of unipolar and bipolar patients. Journal of Affective
Disorders 13, 83-92.
World Health Organization (1977). World Health Organization: Manual of
the International Statistical Classification of Diseases, Injuries, and
Causes of Death (9th ed. Revised, Volume 1). Geneva: World Health
Organization.
World Health Organization (1992). ICD-10. The ICD-10 Classification of
Mental and Behavioural Disorders. Clinical Descriptions and Diagnostic
Guidelines (10th ed. Revised). Geneva: World Health Organization.
Yates, A. J. (1973). Abnormalities of psychomotor functions. In H. J.
Eysenck (Ed.), Handbook of Abnormal Psychology, pp. 261-283. Belfast:
Pitman Medical.
Ylikoski, R., A. Ylikoski, T. Erkinjuntti, R. Sulkava, R. Raininko and
R. Tilvis (1993). White matter changes in healthy elderly persons
correlated with attention and speed of mental processing. Archives of
Neurology 50, 818-824.
Young, M. A., W. A. Scheftner, G. L. Klerman, N. C. Andreasen and
R. M. A. Hirschfeld (1986). The endogenous sub-type of depression: a
study of its internal construct validity. British Journal of
Psychiatry 148, 257-267.
Zimmerman, M., D. W. Black and W. Coryell (1989). Diagnostic criteria for
melancholia. The comparative validity of DSM-III and DSM-III-R.
Archives of General Psychiatry 46, 361-368.
Zimmerman, M., W. Coryell and B. Pfohl (1986a). The validity of the
dexamethasone suppression test as a marker for endogenous depression.
Archives of General Psychiatry 43, 347-355.
Zimmerman, M., W. Coryell, B. Pfohl and D. Stangl (1985). Four definitions
of endogenous depression and the dexamethasone suppression test.
Journal of Affective Disorders 8, 37-45.
Zimmerman, M., W. Coryell, B. Pfohl and D. Stangl (1986b). The validity
of four definitions of endogenous depression: II. Clinical, demographic,
familial, and psychosocial correlates. Archives of General Psychiatry 43,
References 323
234-244.
Zimmerman, M., B. Pfohl, D. Stangl and W. Coryell (1986c). An American
validation study of the Newcastle Diagnostic Scale: I. Relationship with
the dexamethasone suppression test. British Journal of Psychiatry 149,
627-630.
Zimmerman, M. and R. L. Spitzer (1989). Melancholia: from DSM-III to
DSM-III-R. American Journal of Psychiatry 146, 20-28.
Zubenko, G. S. and J. Moossy (1993). Major depression in primary
dementia. Archives of Neurology 45, 1182-1186.
Zubenko, G. S., P. Sullivan, J. P. Nelson, S. H. Belle, F. J. Huff and G. L.
Wolf (1990). Brain imaging abnormalities in mental disorders in late
life. Archives of Neurology 47, 1107-1111.
Zung, W. (1965). A self-rating depression scale. Archives of General
Psychiatry 12, 63-70.
Zung, W. (1980). Affective disorders. In E. W. Busse and D. G. Blazer
(Eds.), Handbook of Geriatric Psychiatry, pp. 338-367. New York: Van
Nostrand Reinhold.
Zweig, R. M., C. A. Ross, J. C. Hedreen, C. Steele, J. E. Cardillo, P. J.
Whitehouse, M. F. Folstein and D. L. Price (1988). The neuropathology
of aminergic nuclei in Alzheimer's disease. Ann. of Neurology 24,
233-242.
Author Index

Aabro, 232 Awad, 245


Abas, 246 Aylward, 248
Abou-Saleh, 138, 144, 146, 274
Abraham, 10 Babor, 39
Abrams, 162, 173, 178, 259 Bachar, 237
Agbayewa, 250 Backman, 246
Agha, 110 Baddeley, 227
Agid, 228 Bailey, 138, 144, 146
Akiskal, 174, 237, 238, 241, 242 Baker, 227
Alavi, 272, 274 Baldessarini, 80, 146
Alazraki, 244, 245 Baldwin, 244, 247, 249, 250
Albala, 40, 139, 140 Ball, 31
Albert, 224, 228 Barber, 80
Alexander, 46, 49, 224, 228-235, 253, Barrett, 39
275 Bartholomew, 58
Alexopoulos, 150, 258, 259 Bartolucci, 248
Allerup, 77 Basford, 61
Almkvist, 246 Bassun, 246
Baxter, 233, 269, 270
Altman, 178 Beats, 244, 248
American Psychiatric Association, 26, Bech, 77
34, 38, 75, 82, 98, 180, 202, 237 Beck, 78, 203, 204
Ames, 250 Behar, 241
Amouyel, 263 Beigel, 173, 177, 178
Ananth, 259 Belendiuk, 232
Andersen, 234, 272 Bell, 225
Anderson, 178, 232 Bellack, 76
Andersson-Lundman, 246 Belle, 241, 247
Andreasen, 1, 20, 57, 60, 82, 252 Bench, 233, 234, 269, 270, 275
Andrews, 22, 39, 82, 161, 179, 249 Bender, 139, 146
Andrezejewski, 234 Bennett, 153-155, 257-259
Angst, 172 Benoit, 79
Aniss, 226, 264 Benson, 229
Antleman, 51 Benton, 152
Apter, 269 Berent, 226, 234, 272
Arana, 146 Berger, 147
Arndt, 265 Bergmann, 250
Asai, 246 Berlin, 274
Asberg, 77 Berman, 234, 271
Attar-Levy, 269, 270 Berrios, 67-69, 79
Austin, 110, 113, 149, 153-155, 192, Berthier, 231, 234
199, 223, 227, 231, 233, 251, 257- Bertrand, 47
259, 267, 269, 270, 273, 274, 279 Berwick, 92

324
Author Index 325
Betley, 272 Buchtel, 234
Bhoopat, 245 Burke, 39
Bianchi, 22, 82, 179 Burns, 227, 228, 250
Biddle, 146 Burvill, 185, 238-240, 249-251
Bierer, 227 Butters, 53, 226-228, 249
Biersack, 274 Byck, 178
Bird, 246, 250 Byrne, 243
Black, 82, 99, 160, 161, 166
Blackburn, 178 Caine, 228, 230
Blass, 259 Calabrese, 39, 264
Blazer, 39-41, 237, 257 Callaway, 248
Blinder, 269 Calloway, 139, 144
Blockx, 147 Calne, 226, 232
Blumenthal, 2 Capdevila, 265
Boer, 241 Cardillo, 241
Boivin, 272 Carmichael, 252, 271
Bolduc, 231 Carney, 1, 22, 33, 76, 86, 98, 136, 138,
Boll, 147 166, 203, 238
Boone, 248, 258 Carpenter, 3, 236
Borbely, 79 Carr, 146, 147
Bosnians, 39, 264 Carroll, 39, 40, 50, 51, 78, 79, 129,
Bouillon, 138, 146 139, 140, 172, 253, 279
Bouthillier, 47 Carson, 240
Bowden, 62, 131 Casper, 62
Bowen, 249 Cassano, 237, 238, 241
Boyce, 18, 22, 23, 28, 32, 63, 84, 90, Cassens, 227
95-97, 99, 103, 110, 113, 115, 121- Caterson, 139, 147
123, 132, 134, 137, 152, 154, 162, Cermak, 228
173, 174, 181, 183-185, 187, 188, Chang, 62
198, 204, 209, 226, 236-238, 241, Channon, 227
246 Charney, 21, 22, 34, 96, 122, 138, 182,
Boyd, 39, 228 223
Boyko, 233, 234, 245, 247, 257, 258 Chase, 245
Bradley, 245 Chiodo, 51
Brand, 227, 228 Christensen, 146
Brandon, 162, 166 Christy, 241
Brandt, 234 Citrin, 246
Brant-Zawadzki, 246 Clark, 269
Breitner, 258 Cleary, 78
Bridge, 78, 146 Cloitre, 79
Bridges, 257, 270 Cochrane, 22
Broadhead, 240 Coffey, 40, 233, 234, 244, 245, 247,
Brockington, 178 248, 253, 257-259, 270
Brodaty, 28, 32, 63, 84, 90, 95-97, 99, Cohen, 227, 275
103, 115, 121-123, 132, 134, 137, Collins, 39
152, 154, 162, 173, 174, 181, 183- Conboy, 248
185, 187, 188, 192, 198, 199, 204, Conwell, 239, 241
209, 226, 236-238, 241, 246 Cooper, 77, 148, 228, 231
Brousseau, 263 Copeland, 250
Brown, 51, 139, 146, 153, 225, 231- Coppen, 138, 144, 146
234, 237, 238, 241, 257, 269 Corder, 47
Brozoski, 231 Cornell, 226
Bruce, 250 Coryell, 82, 93, 99, 122, 138, 139, 144,
Bruna, 265 160, 161, 166, 174, 204, 242
Buchan, 162 Costa, 52, 273
Buchanan, 3, 236 Costello, 1
Buchsbaum, 233, 269, 275 Cosyns, 147
326 Author Index
Cowley, 166 Dolan, 139, 144, 233, 234, 257, 269,
Creasey, 237, 244 270, 275
Cress, 40, 247 Donnelly, 178
Critchley, 274 Dooneief, 225
Cronholm, 227 Doraiswamy, 233, 234, 245, 247, 257,
Cropley, 166 258
Croughan, 78, 173, 178 Dougall, 233, 269, 270, 273, 274
Crow, 162 Downes, 230
Cruickshank, 228 Drayer, 246
Crutcher, 46, 49, 224, 253, 275 Drevets, 252, 271
Cummings, 49, 55, 225, 234, 240, 252, Driesen, 234
257, 279 Dubois, 228
Curran, 11-13, 273, 274 Duggan, 161, 163
Curry, 228 Duke, 248
Duncan-Jones, 243
Dunner, 173
D'Amato, 241 Dupont, 53, 249
D'Hondt, 147 Dwyer, 173
Dallongeville, 263
Daniel, 272
Eagles, 239
Danish University Antidepressant Group Eaton, 39
(DUAG), 77 Ebert, 228, 230
Dannals, 234 Ebinger, 232
Davidson, 138, 144, 235 Ebmeier, 149, 227, 231, 233, 269, 270,
Davies, 147 273, 274
Davignon, 47 Ebrahim, 240
Davis, 62, 69, 131, 235, 244-246 Edland, 240
De Armond, 246 Elkin, 39
De Keyser, 232 Ell, 52, 273
de Leon, 245 Ellard, 83
De Maertelaer, 233, 269 Ellingwood, 258
De Souza, 139, 144 Ellinwood, 40, 233, 234, 245, 257, 258
de Vigne, 40, 139, 140 Elton, 269
Deakin, 166 Emery, 249, 250
Dealy, 182 Emmerson, 238, 239, 249-251
Deaton, 22 Endicott, 78, 136, 174, 180, 242
DeCarli, 245 Endoh, 246
Dejonckheere, 39, 264 Erbaugh, 78
Delecluse, 233, 269, 272 Erkinjuntti, 246
Delgado, 79 Erwin, 271, 273
Delis, 53, 249 Escalona, 247
DeLisi, 233, 269, 275 Eshlemann, 39
DeLong, 46, 49, 224, 228-235, 253, Esiri, 244
275 Evarts, 226
Delvenne, 233, 269 Everitt, 85
Derogatis, 78 Eyers, 28, 32, 63, 84, 90, 96, 97, 99,
Detre, 79, 173, 178 103, 110, 113, 115, 121-123, 152,
Devanand, 148, 269, 274 154, 181, 184, 198, 209, 226, 236
DeLisi, 246 Eysenck, 18, 28, 35, 57, 59
Di Mascio, 227
Diebel, 227, 228 Fahn, 228, 269
Divac, 229 Fairchild, 161
Djang, 40, 233, 234, 244, 245, 247, Farelli, 138
248, 253, 257-259, 270 Farmer, 9, 39, 243
Docherty, 39 Fazekas, 245
Doebereiner, 76 Fedoroff, 265
Doerr, 147 Feinberg, 40, 78, 139, 140
Author Index 327
Feldman, 224 Goldman-Rakic, 228, 229
Feline, 269, 270 Goldsamt, 131
Fenger, 234 Goli, 40, 257
Ferrier, 267 Gonze, 178
Fiester, 39 Goodgold, 245
Fieve, 173 Goodwin, 149, 161, 172, 174, 178, 227,
Figiel, 40, 233, 234, 244, 245, 247, 231, 233, 269, 270, 273, 274
248, 253, 257-259, 270 Gotham, 225, 232, 241
Fleiss, 77 Gram, 77, 146
Fleminger, 226, 264 Granholm, 226-228
Flowers, 228 Granholme, 228
Folks, 248 Grasby, 275
Folstein, 225, 228, 234, 240, 241, 264 Graybiel, 229
Fonagy, 139, 144 Grayson, 18, 35, 63
Fonnum, 229 Greden, 40, 78, 79, 139, 140, 146
Forrester, 265 Greenberg, 239, 241
Foster, 79, 173, 178 Greenwald, 241
Frackowiak, 233, 234, 269, 270, 275, Griffin, 182
276 Grime, 274
France, 40, 257 Gross, 275
Frances, 139, 146, 237, 238, 241 Grove, 57, 60, 82
Francis, 249 Gruenberg, 39
Freedman, 232 Grunhaus, 146
Freidenberg, 241 Grunwald, 274
Friederich, 231 Gudeman, 22
Friedman, 227 Gulley, 224
Friston, 233, 234, 269, 270, 275, 276 Gulmann, 232
Frith, 275, 276 Gunther, 178
Fruchart, 263 Gur, 271-273
Gurland, 77, 237
Gamal, 259 Gurney, 31
Gangestad, 57 Guthrie, 147
Garron, 226 Guze, 233, 269, 270
Garside, 1, 22, 29, 31, 33, 35, 76, 82,
86, 98, 136, 138, 166, 203, 238 Haack, 147
Gaskell, 47 Hadzi-Pavlovic, 22, 23, 28, 32, 35, 63,
Gauthiers, 47 84, 90, 95-97, 99, 103, 110, 113,
Geisler, 265 115, 121-124, 132, 134, 137, 152,
George, 39, 41, 243, 245, 273 154, 162-164, 173, 174, 181, 182,
Georgotas, 138 184, 185, 187, 188, 192, 198, 199,
Gerber, 39 204, 207, 209, 226, 236
Gerfen, 46, 49 Hagman, 269
Gerken, 147 Halaris, 232
Gerner, 233, 269, 270 Hall, 28, 32, 185, 228, 238, 239, 249-
Gershon, 51 251
Gibbons, 62 Hamilton, 22, 29, 75, 78, 97, 238
Giles, 161 Hanin, 62, 182
Gillespie, 12, 16, 34 Hansch, 228, 231
Gilliland, 146, 147 Hapworth, 138
Gillin, 53, 249 Hardy, 269, 270
Giordani, 234, 272 Harrell, 248
Gispen-de Wied, 227 Harrow, 242
Glass, 39 Hart, 46, 226
Glassman, 182, 198 Harvey, 228, 231
Goetz, 18 Harwood, 138, 144, 146
Golden, 63, 85 Haskett, 40, 139, 140, 146
Goldman, 231 Hasselbalch, 234, 272
328 Author Index
Hasselblad, 85 Imber, 39
Haycox, 259 Ionescu, 79, 80, 173, 178
Hay man, 245
Hayward, 228 Jablenski, 39
Hazlett, 233, 269, 275 Jackson, 68, 70, 71, 200
Heald, 230 Jacobs, 264
Healy, 191 Jacoby, 246, 250
Hedreen, 241 Jacomb, 250
Heindel, 53, 249 James, 40, 139, 140
Helzer, 39, 78 Jamieson, 173, 174
Henderson, 12, 39, 227, 243, 250 Jamison, 161, 172, 174, 178
Henry, 232 Jankovic, 18
Herregodts, 232 Janowski, 226-228
Hersen, 76 Jarritt, 273
Hesselink, 53, 249 Jarvik, 227
Hichwa, 234, 272 Jaspers, 72
Hickie, 18, 28, 32, 63, 84, 90, 95-97, Javaid, 62
99, 103, 110, 113, 115, 121-123, Jernigan, 53, 249
132, 134, 137, 152-155, 162, 166, Jipescu, 79, 80, 173, 178
173, 174, 181, 184, 185, 187, 188, Jodar, 265
192, 198, 199, 204, 209, 226,236- Joel, 46, 49
238, 241, 246, 257-259 Johnsen, 272
Hijelmsted, 232 Johnson, 39, 139, 147, 233, 245, 269,
Hill, 224, 240 275
Hillier, 178 Johnstone, 162, 166
Himmelhoch, 76, 178 Jolles, 227
Hirschfeld, 20, 131 Jolley, 250
Hitch, 227 Joo, 47
Hobson, 166 Jordan, 228, 231
Hoffman, 178, 258 Jorge, 265
Holcomb, 233, 269, 275 Jorm, 250
Holden, 138 Josiassen, 228
Hollon, 203, 204 Joyce, 149, 150
Holsboer, 138, 144, 147 Junque, 265
Holt, 39, 259
Holzer, 39 Kahn, 235
Honer, 265 Kakuma, 150, 258, 259
Hope, 35 Kalil, 229
Hopkinson, 241, 246 Kalnin, 245
Horn, 274 Kanaba, 246
Horwath, 39 Kapur, 232, 276
Horwitz, 245 Karasawa, 246
Hough, 39 Karayanidis, 230
Hubain, 233, 269 Karno, 39, 243
Huber, 226, 228, 241 Kasahara, 246
Hudgins, 153, 233, 257 Kaszniak, 226
Huff, 241, 247 Katsikitis, 79
Hughes, 39, 41, 237 Katz, 79, 131, 173, 178
Hunt, 82, 139, 147, 161 Kawamoto, 246
Huppert, 152 Kawashimi, 237
Huret, 269, 270 Kay, 22, 249, 250
Hurwitz, 265, 269 Keller, 174
Husain, 233, 234, 245, 247, 257, 258 Kellner, 246
Huttinot, 148 Kendell, 1, 20, 34-36, 60, 62, 63, 72,
82, 166, 179
Iadarola, 234 Kendler, 39
Ibrahim, 110 Kenn, 228
Author Index 329

Kerns, 248 Lawrence, 80


Kerr, 31, 147 Le Moal, 231
Kerwin, 52 Le Witt, 227, 228
Kessler, 39, 41, 233, 243, 269, 275 Leaf, 250
Khan, 182 Leber, 39
Khatami, 161 Lee, 161, 163, 249
Kiloh, 22, 29, 31, 35, 82, 150, 161, Lees, 226, 228
179, 224, 249 Lehtinen, 234
Kim, 138, 239, 241 Leo, 265
King, 233, 269, 275 Leonhard, 172
Kinoshita, 272, 276 Lesser, 248, 258
Kirkpatrick, 3, 236, 245 Levenden, 230
Kitamuna, 246 Leventhal, 225
Klawans, 226 Levy, 244, 246, 248, 250
Klein, 1, 69, 139, 146 Lewine, 153, 233, 257
Klemm, 274 Lewis, 12-16, 35, 52, 67, 72, 199, 269,
Klerman, 20-22, 33, 34, 237 270
Klonoff, 269 Lhermitte, 228
Knauper, 243 Li, 265
Knight, 153, 263 Liddle, 275, 276
Ko, 235 Lieberman, 259
Kobayashi, 246 Lindesay, 249-251
Kocsis, 139, 146, 237, 238, 241 Linkowski, 39, 138, 146
Koeppe, 272 Linnoila, 172
Kogan, 47 Lipper, 138, 144
Korf, 51, 232 Lishman, 265
Koslow, 62, 131 Lo, 148
Kosugi, 272, 276 Localona, 233, 234, 245, 258
Kouris, 273 Lock, 243
Kovacs, 203, 204 Locke, 39
Kraepelin, 70, 181, 191, 200 Logsdail, 265
Kragh-Sorensen, 77, 146 Lohr, 40, 139, 140
Kramer, 39, 243 Loreck, 240, 241
Kramer-Ginsberg, 241 Loutsch, 237, 238, 241
Kretzschmar, 246 Luszczynska, 240
Krieg, 147
Krishnan, 40, 49, 52, 55, 138, 144, Maas, 62, 131
153, 154, 223, 229, 233-235, 244, Macdonald, 247
245, 247, 249, 251-253, 257, 258, Madeley, 228
263, 267, 279 Maes, 39, 138, 146, 147, 264
Kronfol, 40, 139, 140 Maguire, 146, 147
Kubos, 263 Mahorney, 138, 144
Kuhl, 272 Maier, 138, 144
Kuhs, 173, 178 Mallinson, 13
Kumar, 274 Mancall, 228
Kunzle, 229 Mandel, 179
Kupfer, 76, 79, 173, 178 Mann, 139, 146, 232, 276
Kurlan, 241 Manton, 39, 41
Kuschowski, 228 Mapother, 11, 17
Kuwatani, 272, 276 Marcovinch, 153, 263
Kwentus, 226 Marder, 225
Markel, 234
Laitinen, 232 Marsden, 225, 232, 241
Lakke, 232 Marshall, 245
Lang, 228 Marti-Vilalta, 265
Langston, 233, 269, 275 Martin, 147, 269
Lawler, 166 Martinot, 269, 270
330 Author Index
Martone, 228 Mock, 78
Maser, 242 Modell, 272
Masuda, 246 Moeller, 269, 274
Mathew, 272 Moffoot, 233, 269, 270
Matison, 228 Moller, 274
Matsubayashi, 246 Momose, 272, 276
Mattis, 150, 258, 259 Montgomery, 77, 228, 243
Maudsley, 70, 179 Moossy, 241
Maurer, 82 Moran, 18, 29
Maxwell, 18, 28, 29, 59, 62 Morris, 146, 147, 224, 228, 230, 231
Mayberg, 50, 225, 231, 234, 237, 240, Morris-Yates, 39
241, 252, 257, 262, 264, 269, 270, Mortimer, 228, 231
279 Moss, 228
Mayeux, 225, 228, 241, 274 Mukherjee, 269
Mayr, 139, 146 Mullaney, 31
Mazoyer, 269, 270 Murphy, 35, 60, 92, 173, 177, 178,
Mazure, 239, 241 227, 232, 245, 249-251, 269
Mazziotta, 233, 269, 270 Murray, 52, 149, 161, 163, 227, 231,
McCue, 138 233, 249, 269, 270, 273, 274
McDonald, 59, 60, 100, 153, 166, 233, Muscat, 232
234, 245, 247, 257, 258, 263 Musetti, 237, 238, 241
McDonnell, 234 Myers, 39, 243
McEvoy, 39
McGonagle, 39, 41 Na, 233, 234, 245, 247, 258
McGuffin, 9 Najlerahim, 249
McKeon, 77 Nancroff, 153, 263
McLachlan, 61, 188 Nand, 178
McPartland, 173, 178 Neil, 182
McPherson, 162 Neilson, 22, 82, 161, 179, 249
Meador-Woodruff, 146 Nelson, 21, 22, 34, 39, 96, 122, 138,
Meehl, 57 146, 182, 223, 239, 241, 247
Mehringer, 258 Nemeroff, 40, 233, 234, 245, 247, 257,
Mei-Tal, 239, 241 258
Meltzer, 39, 178, 264 Neshkes, 240
Mena, 258 Nestadt, 39
Mendels, 22, 131 Newberg, 274
Mendelson, 78 Newman, 232
Mendlewicz, 39, 178, 233, 269 Newton, 246, 249
Menzel, 274 Ni Brolchain, 2, 21, 160, 207
Merikangas, 39 Nieber, 76
Merskey, 10 Niederschweiberer, 139, 146
Metcalfe, 138, 144, 146 Nielsen, 146
Meyers, 150, 239, 241, 258, 259 Nishikawa, 272, 276
Miller, 131, 146, 226, 240, 248, 258, Nordahl, 275
259 Norman, 246
Milln, 138, 144, 146 Nouri, 240
Mindham, 228 Nowicka, 241
Minner, 264 Nystrom, 166
Minter, 179
Mirabello, 225 O'Brien, 250
Mirra, 244, 245 O'Carroll, 149, 227, 231, 233, 269, 270,
Mitchell, 28, 32, 63, 84, 90, 95-97, 99,