Melancholia A Disorder of Movement and Mood A Phenomenological and Neurobiological Review PDF
Melancholia A Disorder of Movement and Mood A Phenomenological and Neurobiological Review PDF
Melancholia A Disorder of Movement and Mood A Phenomenological and Neurobiological Review PDF
Edited by
GORDON PARKER and DUSAN HADZI-PAVLOVIC
School of Psychiatry, University of New South Wales, 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
ISBN 0-521-47275-X
Vll
viii Contributors
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
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).
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)
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.
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.
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.
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.
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).
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.
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
////////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
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.
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
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
57
58 D. Hadzi-Pavlovic
(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.
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
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
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.
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
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:
Percentage Factor
affirming Odds loadings
Symptom PD/ED ND/RDf ratio I II
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.
Item probabilities
Factor in LCA
Mental state sign loading! Class A Class B
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.
Factor
Sign I II III
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.
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.
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
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.
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.
(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.
20 25
(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.
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.
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.
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
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.
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
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
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).
2 x 2 chi-squared statistics: *P < 0.05; **P < 0.01; ***P < 0.001.
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.
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
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
N= 5
100-
Observed
90- Expected under logistic model
80
60-
50-
40-
30-
20-
10-
0-
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
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).
(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).
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
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
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
SYMPTOM scores
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.
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
§ -,
dence that, for those receiving antidepressants, a higher CORE score was
associated with an increased probability of recovery.
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*
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
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.
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%).
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
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.
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.
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.
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.
Clinical diagnosis n n
Psychotic depressives 34 100% 0 0%
Endogenous depressives 11 12% 80 87%
Total 45 36% 80 64%
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.
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).
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%
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.
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
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
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).
211
212 K. Wilhelm
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.
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.
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.
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.
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).
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.
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.
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
237
238 H. Brodaty
<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
<40 (N = 116)
40-59 (N = 83)
15 - 60 + (N = 52)
o> 10 -
5 -
J
0
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.
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.
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.
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
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.
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
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.
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.
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
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 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.
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
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
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
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
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
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
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
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
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
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
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
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
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
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.
Circle whether the mental state examination was undertaken near to or at the
nadir (most severe period) of the episode.
Nadir: Yes No
Retardation: 2 + 3 + 6 + 10 + 13 + 15 + 17 =
Agitation: 5 + 9 + 11 + 14 + 18 =
292
References 293
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
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
Zee, 234
Zhao, 39
Zimmerman, 82, 93, 99, 122, 138, 139,
144, 160, 161, 166, 204
Zisook, 237
Zola, 227
Zubenko, 241, 247
Zubrick, 228
Zung, 78, 97, 138, 144, 237
Zweig, 241
Subject Index
335
336 Subject Index