Author’s Accepted Manuscript

Atypical Depression and Non-Atypical Depression:

Is HPA Axis Function a Biomarker? A Systematic
Review

Mario F. Juruena, Mariia Bocharova, Bruno

Agustini, Allan H. Young
www.elsevier.com/locate/jad

PII: S0165-0327(17)30691-2
DOI: https://doi.org/10.1016/j.jad.2017.09.052
Reference: JAD9260
To appear in: Journal of Affective Disorders
Received date: 2 April 2017
Revised date: 11 July 2017
Accepted date: 26 September 2017
Cite this article as: Mario F. Juruena, Mariia Bocharova, Bruno Agustini and
Allan H. Young, Atypical Depression and Non-Atypical Depression: Is HPA
Axis Function a Biomarker? A Systematic Review, Journal of Affective
Disorders, https://doi.org/10.1016/j.jad.2017.09.052
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
 Atypical Depression and Non-Atypical Depression: Is HPA Axis Function a

Biomarker? A Systematic Review

Mario F. Juruena1,2*, Mariia Bocharova1, Bruno Agustini1, Allan H. Young1

1
Centre for Affective Disorders, Department of Psychological Medicine, Institute of
Psychiatry, Psychology and Neuroscience, Biomedical Research Centre (BRC) at
South London and Maudsley NHS Foundation Trust (SLaM) and King’s College
London.UK
2
Department of Neuroscience and Behavior, School of Medicine of Ribeirao Preto,
University of Sao Paulo, Sao Paulo, Brazil;

*Correspondence to: PO72, De Crespigny Park-Denmark Hill, London SE5 8AF, UK.
[email protected]

Abstract

Background
The link between the abnormalities of the Hypothalamic-pituitary-adrenal (HPA) axis and
depression has been one of the most consistently reported findings in psychiatry. At the
same time, multiple studies have demonstrated a stronger association between the in-
creased activation of HPA-axis and melancholic, or endogenous depression subtype. This
association has not been confirmed for the atypical subtype, and some researchers have
suggested that as an antinomic depressive subtype, it may be associated with the opposite
type, i.e. hypo-function, of the HPA-axis, similarly to PTSD. The purpose of this systematic
review is to summarise existing studies addressing the abnormalities of the HPA-axis in
melancholic and/or atypical depression.
Method
We conducted a systematic review in the literature by searching MEDLINE, PsycINFO,
OvidSP and Embase databases until June 2017. The following search items were used:
"hypothalamic-pituitary-adrenal" OR "HPA" OR "cortisol" OR "corticotropin releasing hor-
mone" OR "corticotropin releasing factor" OR "glucocorticoid*" OR "adrenocorticotropic hor-
mone" OR "ACTH" AND "atypical depression" OR "non-atypical depression" OR "melanchol-
ic depression" OR "non-melancholic depression" OR "endogenous depression" OR "en-
dogenomorphic depression" OR "non-endogenous depression". Search limits were set to
include papers in English or German language published in peer-reviewed journals at any
period. All studies were scrutinized to determine the main methodological characteristics,
and particularly possible sources of bias influencing the results reported.
Results
We selected 48 relevant studies. Detailed analysis of the methodologies used in the studies
revealed significant variability especially regarding the samples’ definition comparing the
HPA axis activity of melancholic patients to atypical depression, including healthy controls.
The results were subdivided into 4 sections: 1) 27 studies which compared melancholic OR
endogenous depression vs. non-melancholic or non-endogenous depression or controls; 2)
9 studies which compared atypical depression or atypical traits vs. non-atypical depression
or controls; 3) 7 studies which compared melancholic or endogenous and atypical depres-
sion subtypes and 4) 5 studies which used a longitudinal design, comparing the measures of
HPA-axis across two or more time points. While the majority of studies did confirm the asso-
ciation between melancholic depression and increased post-challenge cortisol levels, the
association with increases in basal cortisol and basal ACTH were less consistent. Some
studies, particularly those focusing on reversed vegetative symptoms, demonstrated a de-
crease in the activity of the HPA axis in atypical depression compared to controls, but the
majority did not distinguish it from healthy controls.
Conclusions
In conclusion, our findings indicate that there is a difference in the activity of the HPA-axis
between melancholic and atypical depressive subtypes. However, these are more likely ex-
plained by hypercortisolism in melancholia; and most often normal than decreased function
in atypical depression. Further research should seek to distinguish a particular subtype of
depression linked to HPA-axis abnormalities, based on symptom profile, with a focus on
vegetative symptoms, neuroendocrine probes, and the history of adverse childhood events.
New insights into the dichotomy addressed in this review might be obtained from genetic
and epigenetic studies of HPA-axis related genes in both subtypes, with an emphasis on the
presence of vegetative symptoms.

Keywords: hypothalamic-pituitary-adrenal axis, cortisol, atypical depression, me-

lancholic depression, endogenous depression.

Introduction

Major depression is undoubtedly one of the major healthcare issues in the

21st century According to the latest WHO report on 23 February 2017, depression is
now ranked as the single largest contributor of years lived with disability worldwide,
and the major contributor to the global burden of disease. (WHO, 2017)

Stress response system abnormalities in Depression

The most robust and consistent finding in major depression so far has been its link to
the abnormalities of the stress response system. The stress response system is a
complex, multilevel mechanism largely dependent on feedback regulation. It relies
on two main elements - the autonomic stress response which exerts immediate ef-
fects when the organism is faced with physiological or psychological stressors; and
the impact on the hypothalamo-pituitary-adrenal (HPA) axis. A detailed account of
the structure and physiology of both components of the stress response system is
thoroughly given by Ulrich-Lai & Herman (2009) and is outside the scope of this re-
view. However, to understand the nature of the abnormalities which will be discussed
further, it is essential to highlight the key characteristics of the HPA-axis.
In response to stressful stimuli, the suppression of the subgenual prefrontal cortex
and the activation of amygdala lead to the stimulation of the autonomic sympathetic
axis, and the HPA axis (Dioro et al. 1993; Phelps and LeDoux, 2015; Gold, 2015).
The autonomic sympathetic axis is responsible for the most rapid response, and acts
via the secretion of epinephrine by the adrenal glands; the HPA axis is activated mi-
nutes after the epinephrine surge, and represents a cascade of events starting with
the secretion of the corticotropin releasing factor (CRF, also known as corticotropin-
releasing hormone, or CRH) from the paraventricular nucleus of the hypothalamus
into the portal circulation, which stimulates the synthesis and release of adrenocorti-
cotropic hormone (ACTH) by the pituitary. The ACTH further stimulates the synthe-
sis and release of the glucocorticoid hormone cortisol by the adrenal cortex. Gluco-
corticoids are known to exert a range of functions, such as promoting gluconeogene-
sis, catabolic and antianabolic activity, suppression of innate immunity in immune
organs, insulin resistance and a prothrombotic state. The key role of glucocorticoids
consists in maintaining homoeostasis in response to stress (Juruena, 2014).
The HPA-axis exerts its feedback through two major types of receptors: the gluco-
corticoid (GR) and the mineralocorticoid (MR) receptors. MRs have higher affinity to
cortisol which results in their higher occupancy even at basal cortisol concentrations;
at the same time, they are less specific and bind with both cortisol and aldosterone.
GRs, on the other hand, bind more specifically to cortisol, yet respond to higher con-
centrations than MRs MR’s seem to be the ones that regulate cortisol feedback du-
ring acute or normal stress. However, during severe or prolonged stress, GR’s come
into action (De Kloet et al.,1998).
The abnormalities of stress response system in affective disorders have been im-
plied in several hundred studies (Stetler et al. 2013). However, accumulating evi-
dence suggests that the presence, and type of, HPA-axis abnormalities may vary
across various subtypes of depression (Gold et al. 2015; Porter and Gallagher
2006). Some studies have shown a robust association with HPA-axis overactivity
with more severe or endogenomorphic forms of depression such as melancholic de-
pression or psychotic depression (Nelson and Davis 1997).At the same time, in post-
traumatic stress disorder, an enhanced HPA negative feedback was described
(Yehuda et al. 1991). Some studies have hypothesised that atypical depression, un-
like the melancholic subtype and similar to PTSD, is characterised by hypocortiso-
lism and enhanced negative feedback. However, whether it is fair to claim there
exists such a dichotomy, is not clear at the moment.
 The aim of the current article is to review existing literature addressing the function
of the HPA-axis in melancholic and atypical depressive subtypes. Evaluate whether:
a) there is a significant difference between the two subtypes in terms of the activity of
the HPA-axis; b) whether there is enough evidence to suggest that the HPA-axis is
overactive in depression with melancholic features; c) whether there is sufficient evi-
dence to indicate that the HPA-axis in depression with atypical features is hypoac-
tive.

Melancholic vs atypical depression: a historical perspective on subtype defini-

tion and boundaries.

Considering the studies which have addressed HPA-axis abnormalities in either of,
or both, melancholic and atypical subtypes, it is important to take into account that
over the recent decades, approaches to identify them have been changing, and even
today, appropriate criteria defining both subtypes remain a matter of debate.
A specifier introduced in DSM-III (1980), depression with melancholic features re-
presents a subtype of depression clinically characterized by a distinct pattern of low
mood, anhedonia, lack of reactivity to positive events, loss of appetite and weight,
insomnia, loss of libido and diurnal mood variations. Although depression with «me-
lancholic features» has been validated extensively (Shotte et al, 1997; Juruena et al
2011; Parker et al, 2015), there is still little agreement among researchers regarding
the particular set of features that define it (Maes et al., 1992; Leventhal et al, 2005 ;
Fink et al, 2007; Parker et al, 2013). Besides, the various diagnostic measures used
for identifying melancholic depression have shown a considerable degree of incon-
sistency, as exemplified by the comparison of Research Diagnostic Criteria (RDC)
endogenous depression definition, Newcastle scale endogenous depression defini-
tion, and DSM-III and DSM-IV diagnoses of melancholic depression (Rush et al.,
1994 ; Coryell, 2007 Orsel et al., 2010).
Nevertheless, it is of high importance for this review to point out that the sub-
sample identified by researchers as non-melancholic can be highly heterogeneous
and represent various diagnostic entities, such as neurotic/reactive, atypical, non-
differentiated depression, characterological depression (Fink and Taylor, 2007).
Therefore, although atypical depression is indeed greatly antithetic to melancholic
depression, it is not the same as non-melancholic depression, and for the purpose of
this review, unless specified as having features known to constitute the atypical sub-
type, non-melancholic subsamples of reviewed studies will not be not considered
atypical.
Atypical depression had been recognized by some researchers as a depressive
subtype since the 1960s. However, although addressed extensively in the context of
differential responsiveness to pharmacological treatment, it had not been included in
official DSM diagnostic criteria until 1994, following the formulation of Columbia aty-
pical depression criteria (Quitkin et al. 1993). The validity of the atypical specifier
has been demonstrated by some studies, that largely rests on data from psycho-
pharmacological research - an approach introduced by Klein and according to which,
the differential response to biological treatment represents different pathophysiology
(Klein, 1989) and genetic-epidemiological studies which indicated that atypical de-
pression subtype is genetically distinct from typical ones, which represents its etiolo-
gical insularity (Kendler et al., 1996; Sullivan et al..1998). At the same time, the pre-
cise definition of «atypical depression» remains a matter of debate. According to
DSM-IV diagnostic criteria (“atypical features” specifier), the disorder is primarily cha-
racterized by mood reactivity and 2 or more of the following symptoms as predomi-
nant features in patients with major depression or dysthymic disorder: overeating,
oversleeping, “leaden paralysis,” and interpersonal rejection sensitivity. The rele-
vance of mood reactivity to the subtype, as well as its relationship with other symp-
toms, has been questioned. The studies cited above as providing evidence for the
validity of the subtype, in fact, did not include the «mood reactivity» criteria. In a stu-
dy by Posternak et al. (2001) mood reactivity did not show any association with any
other psychopathological features of the proposed subtype. Similar results were re-
ported by a few other researchers (Parker et al., 2002). A much more robust asso-
ciation was shown for the component of the symptom definition described as «re-
versed vegetative symptoms», namely hyperphagia and hypersomnia, and some of
the studies selected for this review have focused on these characteristics rather than
the DSM-defined specifier (Benazzi, 2002). In two studies of treatment response,
Stewart et al. (2007) demonstrated a significant contribution of factors such as age
of onset (early vs late) and chronicity (chronic or remitting course) to the correlation
between atypical subtype and response to MAOI treatment.. Some authors proposed
a reappraisal of the DSM criteria for an atypical subtype to reduce the number of
symptoms besides mood reactivity to one, and to include the onset of dysphoria be-
fore 20 and chronic course as additional criteria. However, DSM-5 saw no amend-
ments to the subtype criteria (Stewart et al. 2007).
Therefore, the definition of the atypical subtype appears even more vague than that
of melancholic subtype, and data of all biological studies should be interpreted with
account of the characteristics of the phenotype. Besides, in the majority of studies,
as well as in the case with non-melancholic depression, the non-atypical descriptor
does not necessarily identify melancholic depression, therefore, unless specified
otherwise, non-atypical depression is considered as MDD not matching particular
subtype criteria.

Methods

We conducted a systematic review of the literature by searching Ovid MEDLINE(R)

Epub Ahead of Print, In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R)
Daily and Ovid MEDLINE(R) 1946 to Present, PsycINFO, Journals@Ovid Full Text
and Embase databases using the Ovid platform.

The following search items were used:

"hypothalamic-pituitary-adrenal" OR "HPA" OR "cortisol" OR "corticotropin releasing

hormone" OR "corticotropin releasing factor" OR "glucocorticoid*" OR "adrenocorti-
cotropic hormone" OR "ACTH" AND "atypical depression" OR "non-atypical depres-
sion" OR "melancholic depression" OR "non-melancholic depression" OR "endoge-
nous depression" OR "non-endogenous depressive» Search limits were set to in-
clude papers in English or German language, published in peer-reviewed journals at
any period. Fig. (1) shows details of the search strategy.

The initial search yielded 9556 results in total, which was comprised of: 3192 results
in the Embase database, 2029 in all MEDLINE databases, 2344 in PsycINFO data-
base, and 1991 in Journals@ Ovid Full-Text database. Because the combined num-
ber of articles across all selected databases (9556) exceeded the 6000 limit allowing
for deduplication, Further deduplication limited the number of articles to 3061.

Next, the filters «Human», «Adult» and «Depression» were applied, leaving 570 arti-
cles for title/abstract screen. For the full-text screen, we included original studies as-
sessing the functional elements of the HPA-axis in samples including at least one of
the following phenotypes: melancholic depression, endogenous depression, en-
dogenomorphic depression, atypical depression, their characteristic traits (such as
«reversed neurovegetative symptoms»), and a comparison group.

After the title/abstract screen, 247 articles were retrieved for a full-text screen.

The criteria for inclusion in the review were:

a) The inclusion of adult patients diagnosed with major depressive episode ac-
cording to DSM or ICD operational criteria;

b) An explicit description of criteria for melancholic and/or atypical depressive sub-

type identification in patients/ or precise description of symptom sets characteristic of
melancholic/atypical subtype (e.g. reversed neurovegetative symptoms)

c) The evaluation of the levels of basal and/or post-challenge cortisol in the blood,
saliva, urine or CSF, of basal and/or post-challenge ACTH in blood, and of basal
and/or post-challenge CRF in blood or CSF.

Studies were excluded if:

a) They had no comparison group (e.g. studies in a small sample of melancholic pa-
tients before and after TMS, however, studies which compared the subtypes be-
tween each other without healthy controls, were included)

b) They addressed other neuroendocrine outcome measures (e.g. NE, Prolactin,

Vasopressin) - unless they also assessed HPA-axis measures. Here it is important to
acknowledge that the function of the HPA-axis is largely dependent on, and interre-
lated with, other endocrine factors and the immune system. However, given the fo-
cus of this review, measurements other than those directly related to the HPA-axis
were not considered.
The details of the literature search have been structured using the Prisma Flowchart
in Fig.1
Fig.1 PRISMA Flow Diagram

Results

The final stage of the search process yielded 48 articles in general. The studies were
subdivided into four groups:
Group 1. Studies (n=27) which compared «melancholic» (or other definitions) de-
pression with depression not matching any of the definitions for melancholia, or a
control group
Group 2. Studies (n=9) which compared «atypical» (or other definitions) depression
with depressions not matching any of the definitions for atypicality, or a control
group.
Group 3. Studies (n=7) which compared «melancholic» (or other definitions) de-
pression with «atypical» (or other definitions) depression.
Group 4. Studies (n=5) which employed a longitudinal design.
The results for each section are summarized in Tables 1-4, respectively

Melancholic studies

The group comparing various definitions of melancholic depression with non-

melancholic depression or controls included 27 studies. For clarity and simplicity, we
suggest that within the group, these studies further be sub-grouped according to the
measurement of HPA-axis activity used. However, since quite a few of the studies
addressed more than one potential measure of HPA-axis activity, some of them may
be mentioned several times both in further text and the tables.
The majority of studies assessing the function of the HPA-axis in melancholic de-
pression have focused either on the rates of suppression following Dexamethasone
Suppression Test (17 studies) or basal cortisol in either blood, CSF, or urine (16 stu-
dies). Those are followed by basal ACTH (3 studies), basal CRH (CRF) (2 studies),
oCRH challenge (1 study), ACTH stimulation (1 study), fenfluramine challenge (2
studies), Dex/CRH (1 study).

Dexamethasone suppression test

 Some of the earliest studies focusing on DST as a potential diagnostic test for me-
lancholic depression were performed by B.Carroll et al. This review focuses on the
1981 study which lasted for 6 years and included 368 patients. In this study, the au-
thors addressed not only the issues of the specificity and sensitivity of DST but also
proposed the standard procedure, since the methodologies employed by various
researchers vary markedly in the dose of administered dexamethasone, the time of
measurement, as well as the threshold for «suppression». The authors concluded
that the optimal balance between the measures of sensitivity, specificity and dia-
gnostic confidence of DST for melancholic depression would be at the threshold of
the non-suppression set at 5 μg/dL, the dose of dexamethasone set at 1 mg, and the
measurements were taken at 4 pm and 11 pm. The latter had been demonstrated in
their previous studies which indicated that the abnormalities of the HPA-axis are sub-
tle in depression and therefore patients may still be capable of normal suppression in
the morning hours but fail to do so later in the course of 24 hours post-challenge
(Carroll et al., 1982, further confirmed by Rubin et al., 1987). Similar findings were
also observed by Rubin et al., (1987), in a study which will be discussed further. In
general, the sensitivity (defined here and further as the proportion of melancholic pa-
tients with abnormal DST results) across all values varied from 39% to 53%, the
specificity (defined here and further as the percentage of melancholic subjects in
whom normal results were observed) ranged from 85% to 97%, and the diagnostic
confidence, i.e. the proportion of abnormal test results that were true-positive for me-
lancholia, from 83% to 93 %. These results appear rather convincing, but an essen-
tial aspect of their findings is that for the definition of melancholia, they used their
operational criteria.
Banki et al. (1986), apart from other hormonal challenges, investigated DST res-
ponses in female patients diagnosed with DSM-III defined melancholic depression,
and reported that abnormal DST response was observed in 67% of melancholic pa-
tients, which significantly exceeded the rates for both other psychiatric disorders and
controls (p<0.05). Evans et al. (1987), reported slightly higher rates of non-
suppression for DSM-III melancholic depressed patients - 78%, however, it is of note
that patients with psychotic depression demonstrated an even higher proportion of
non-suppression (95%), which significantly exceeded that in the melancholic group
even when non-suppression was defined as >10 μg/dl or >15 μg/dl.
 The issue of the threshold for non-suppression was addressed in a study by
Winokur et al. (1987). The authors looked at the extremities of the DST response:
having tested 423 patients with a range of affective diagnoses, they split the results
into two categories: strong suppressors defined as having post-dec cortisol levels
below 1.5 μg/dl, and strong non-suppressors whose post-dec cortisol was > 6 μg/dl.
The main results were that melancholic symptoms (defined by DSM-III criteria) were
significantly associated with non-suppression (24% vs. 9% suppressors, p=0.01). At
the same time, the diagnosis of secondary depression was significantly associated
with suppression (38% vs. 19%, p=0.025). It is notable, however, that although there
existed a significant association of non-suppression with melancholia, there were
indeed only 24% of NS among all melancholic patients, which may indicate that the
threshold of 6 mg/dl could be too high to yield sensitive results. (Winokur et al.,
1987).
Contributing to the debate about the correct classification, or the ability of a certain
definition of melancholia to identify non-suppressors, Peselow et al. (1992) demons-
trated that patients who met both DSM-III criteria for melancholia and RDC criteria
for endogenous depression, showed higher rates on non-suppression than those
with «neither» subtype and controls, but not compared to those with «either of» the
subtypes (Peselow et al., 1992). Paslakis et al. (2011) also used DSM-IV (SCID-
validated) criteria for melancholia, and although they did demonstrate a significant
difference in suppression levels between melancholic depressed patients and con-
trols, both the mean effect size and the sensitivity of the test proved low (Paslakis et
al., 2011). Interesting results, although in a very small number of patients (n=5 pa-
tients with "rapid improvement") were obtained by Barocka et al.,(1987). They
showed that the clinical course of "rapid improvement» was significantly associated
with normal suppression in the DST even in endogenous patients (p = 0.04). They
suggested that by eliminating those patients who show a rapid improvement shortly
after the test, the test sensitivity for endogenous depression could be increased by
about 10%, while the specificity remains constant (Barocka et al., 1987).
Rubin et al. (1987) addressed the response to DST (with a threshold set at ≥ 3,5
μg/dL), in 40 patients with a definite RDC diagnosis of endogenous depression vs.
40 age-matched controls. Interestingly, even with this low a threshold, they only ob-
served non-suppression in 15 of the 40 depressive patients, which, however, was
significantly different from controls, yielding a sensitivity of 38% and a specificity of
88%. Their key observation was the melancholic patients who differed from each
other of DST suppression rates were also different on some other HPA-related cha-
racteristics. At the same time, the suppressors did not differ from control subjects in
any of the measures.
Related findings were reported by Amsterdam et al. (1989) who performed a com-
plex study in two consecutive samples, employing a range of challenge tests. Tests
other than DST will be discussed in the relevant section. Consistent with the results
obtained by Rubin et al.,(1987) they found that the subgroup of melancholic patients
who were DST non-suppressors demonstrated larger mean cortisol values for all of
the response measurements compared to the other patient subgroups or healthy
controls, and to all the other patient groups combined. Also, the melancholic DST-NS
subgroup showed a skew toward larger overall cortisol response values, rather than
the standard distribution of the healthy controls, indicating that this group of patients
had members with particularly enhanced adrenocortical responsiveness. Besides,
maximum cortisol response to ACTH was significantly decreased after treatment in
the MEL/DST-NS group (p=0.04). Consistent with the conclusions made by Rubin et
al., the authors inferred that the subgroup of DST-NS patients with melancholic fea-
tures might represent a diagnostically homogeneous subpopulation who are more
likely to demonstrate endocrine abnormalities at several sites within the HPA axis
(Amsterdam et al. 1989).
However, there were studies that did not show a strong association between
the diagnosis of endogenous depression and endogenicity. Hubain et al. (1996) per-
formed DST in a large sample of 155 Newcastle Endogenous Depression Diagnostic
Index (NEDDI) - defined endogenous patients vs. a similar number of non-
endogenous patients. Authors failed to find any association, at a suppression
threshold of 50 μg/l (i.e. 5 μg/dL). (Hubain et al., 1996). Berger et al. reported results
both for the common ≥ 5 μg/dl threshold and for ≥ 8 μg/dl ( a result that yielded 2.7%
in a healthy sample). In their study, in three separate patients samples including the
diagnosis of a) endogenous (+schizoaffective) depression, b) neurotic and situational
depression and c) other psychiatric diagnoses, a positive DST failed to yield predic-
tive value over 40-60%, indicating that only one of two patients with positive DST
would suffer from endogenous depression. For subtyping, authors used three ins-
truments: ICD-8, RDC and Newcastle Scale (Berger et al., 1984).
 Particular attention should be paid to studies which focused not, or not only, on
subtypes as a whole, but also on individual symptoms which reflect the more «biolo-
gical» manifestations of depression. For instance, Miller and Nelson (1987) investi-
gated the DST response in 95 depressed inpatients, 45 of them melancholic accor-
ding to DSM-III criteria (although they also used RDC criteria which did not contri-
bute much to the variance apart from a slight difference in p-values). However, they
assessed the association of individual symptoms, not subtypes as a whole, with
suppression rates. They demonstrated that with a threshold for non-suppression of 7
μg/dL (which is obviously higher than in general), the four factors most strongly as-
sociated with DST non- suppression were initial insomnia, loss of sexual interest,
agitation, and weight loss. Factors associated significantly yet with a small effect size
were also retardation and when cortisol levels were assessed as a continuous va-
riable, ruminative thinking and midnight awakening. It is, nevertheless, notable that
the four strongly associated factors were only shown to account for 24% of the va-
riance. In a study by Berger et al., despite low predictive value observed, weight
loss was shown to enhance the rate of abnormal DST results in psychiatric in-
patients, regardless of their diagnostic classification (Berger et al., 1984).
Casper et al. (1987), also focused on neurovegetative symptoms of depression
rather than defined subtypes. They assessed basal and post-DST cortisol levels in
28 patients diagnosed as having MDD and presenting with either marked weight
loss, appetite loss, or both. Although both weight loss and appetite loss were signifi-
cant predictors of elevated basal cortisol levels, non-suppression, defined as cortisol
levels above or equal to 6 μg/dL, was only significantly associated with weight loss.
Interestingly, as will be discussed in the relevant section, in another study comparing
MDD patients with hypersomnia and/or increased appetite/ weight with MDD patients
without these symptoms or controls did not reveal a strong association between re-
versed vegetative symptoms and HPA-axis function: the most significant association
was demonstrated for DST non-suppression in MDD patients without hypersomnia or
weight/appetite increase vs. the opposite or controls (Casper et. al., 1988). Another
study focusing on individual symptoms was that by Maes et al. (1989), who distin-
guished between vital vs. non-vital symptom clusters instead of MDD subtypes, and
they as well demonstrated that higher levels of cortisol post-DST were associated
with sleep and appetite disturbances.
 At the same time, Orsel et al. (2010) performed a cluster analysis and identified an
endogenous cluster which differed from non- endogenous on the following characte-
ristics: anorexia/weight loss; diurnal variation, depressed mood, loss of energy, early
morning awakening, loss of interest, suicidal ideation, distinct quality of mood, cogni-
tive disturbances, psychomotor disorders, psychotic symptoms, non-reactivity, fee-
lings of guilt, and sleep disorder. However, of the 14 SCID-I items, only six factors -
early morning awakening, distinct quality of mood, feelings of guilt, non-reactivity,
suicidal ideation, and psychomotor disorders - were significant discriminators bet-
ween the clusters. The clusters showed a high degree of correlation with DSM-IV
melancholic and non-melancholic depression, respectively, although they were not a
complete 100% match. The authors were able to demonstrate a significant difference
in basal cortisol levels between endogenous and non- endogenous clusters, as well
as in rates of DST suppression (≥ 3.5 μg/dl threshold), however, not in post-DST cor-
tisol measured as a continuous variable. Moreover, rather unusually, in studies dis-
cussed above, only mood reactivity as a single symptom differed significantly bet-
ween suppressors and non-suppressors (i.e. non-reactivity correlated with non-
suppression; Orsel et al., 2010).
Finally, important observations regarding the role of phenotype in HPA-axis as-
sessment in depression can be inferred from a study by Halbreich et al., 1989, who
compared patients with RDC Endogenous Depression and DSM-III PTSD with RDC
MDD-ED patients alone and controls. They demonstrated that the presence of a dia-
gnosis of PTSD determined a significant difference in DST response in patients: all
PTSD-ED patients proved to be suppressors (not different from controls – however,
suppression threshold wasn’t specified in the article, except for mean values of post-
DST cortisol which were at mean of 3,72 in MDD-ED patients compared to 0,96 in
PTSD-ED patients, p=0.01) and had lower basal plasma cortisol levels compared to
MDD-ED patients (Halbreich et al., 1989).

Other challenge tests

Amsterdam (1989) attempted to assess several levels of the HPA axis and their ab-
normalities in hormonal response in depressive patients. In their paper, they referred
to 2 studies which used ACTH as a hormonal challenge. The first one compared 16
patients with a depressive disorder (of which, 9 had melancholic features) and 11
healthy controls. They found no difference in cortisol levels at baseline between
groups. However, after ACTH administration, the depressive group had a larger in-
crease in cortisol concentrations than controls (p<0,02). Further, the authors tried to
replicate these results in a larger sample of 72 depressed patients (of which 51 had
melancholic features) compared to 34 healthy controls and were unable to find the
same results. In contrast, they found that both at baseline and after ACTH admin-
istration, cortisol levels did not significantly differ between groups. The same article
cites a study with ovine CRH challenge. They administered oCRH to 26 depressed
patients (of which 14 had melancholic features) and compared to 11 healthy con-
trols. They found no difference between groups in cortisol levels at baseline and after
challenge; nonetheless, depressive patients as a group had lower ACTH response
compared to controls. The effect was larger in patients with melancholic features
(Joseph-Vanderpool et al 1991).

The study by Mitchell et al., (1990) tried a different challenge using the serotonin ag-
onist Fenfluramine 60 mg but found no difference between groups of “endogenous”
vs. “non-endogenous” depressive patients. The 30 patients in the study were evalu-
ated with four different types of criteria (DSM-III, ICD 9, RDC, Newcastle criteria) to
divide subtypes; however, they did not found differences in post-challenge cortisol
levels between any of the groups.

Another study that used a similar compound, in this case, d-Fenfluramine 30 mg, al-
so failed to show any significant difference. The study compared 23 depressive pa-
tients with 16 healthy controls and found that both groups had the same levels of
cortisol after challenge (O'Keane & Dinan, 1991).

In 2011, Paslakis (Paslakis et al., 2011) compared 3 ways of assessing HPA axis
and their relevance for detection of depression. One of those was the Dex/CRH test.
They found that the test had low levels of sensibility (30,8%) and moderate specifici-
ty (78,8%) being the worst marker between the three compared (DST, 24h cortisol).

Basal measurements

Basal cortisol
The search identified 19 studies which compared the basal cortisol levels (either as
part of further DST or separately) among patients diagnosed with melancholic de-
pression and matched controls and/or non-melancholic depressed patients.

Rubin et al. (1987) demonstrated that the elevation of basal cortisol when meas-
ured continuously over 26 hours correlated most significantly with the DST status of
the patients: i.e. patients that were non-suppressors had elevated 24-hour cortisol
and nocturnal cortisol nadir compared to both controls and patients who were sup-
pressors. This was shown in a sample of RDC-examined definite endogenous pa-
tients (Rubin et al., 1987).

Further evidence of the increase in basal cortisol levels between melancholic de-
pressed patients and controls/non-melancholic patients comes from studies by Gue-
chot et al. (1987), Wong et al. (2000), Paslakis et al. (2011), O’Keane and Di-
nan(1991). Guechot et al. performed a simple one-time salivary test of cortisol at 11
pm in patients diagnosed as «primary» depressive using DSM-III and Saint-Louis
criteria, compared to both secondary depressives (p<0.05) and controls (p<0.02).
They also reported high sensitivity (62,5%) and specificity (75% vs. secondary de-
pressives, 90% vs. controls) rates of the test for the identification of primary depres-
sive patients when the cut-off was set at 3.45 nmol/l (Guechot et al., 1987). A study
by Wong et al. (2000), addressed multiple measures including basal plasma cortisol
in patients with a DSM-II-R and RDC-defined melancholic subtype of depression and
controls, concluded that mean 30-h plasma cortisol levels in depressed patients
were significantly higher vs. controls. Paslakis et al. (2011), together with DST and
Dex/CRH, which will be discussed, further, performed a comparison of Basal cortisol
levels in melancholic patients defined using DSM-IV criteria vs. healthy controls.
They confirmed that diurnal basal cortisol secretion as measured throughout 24
hours, is significantly higher than in controls, as measured both by cortisol profile
graphs and area under the curve (p=0.001 and p<0.01, respectively). The maximum
of variation was observed at 11.30 and 14.00. In this study, basal cortisol was the
most sensitive marker for melancholic depression than DST, and at the time interval
from 10.00 to 12.00 yielded results with optimal sensitivity (83.3%) and specificity
(87.9%). O’Keane and Dinan (1991) study also reported elevated baseline cortisol in
patients diagnosed with DSM-III-R major depression and Newcastle scale criteria for
endogenicity. They showed a significant increase in basal cortisol in patients com-
pared to controls subjects. CORE measure was also used by Liu et al.(2016) in the
definition of melancholic depression. They performed an analysis of 228 blood meta-
bolic markers in 21 melancholic patients (as well as 58 patients matching criteria for
«anxious depression» defined by one or more comorbid anxiety disorders on M.I.N.I,
and 100 controls). One of the important outputs of this study was the confirmation
that melancholic depression can represent a more biologically distinct subtype of de-
pression. Regarding HPA-axis activity, increases in basal cortisol levels were signifi-
cant for the melancholic group, as well as increases in other metabolites in the hor-
mone biosynthesis pathway (androstenedione and corticosterone; Liu et al., 2016).

Michopoulos et al. (2008) studied whether elevated HPA-axis function was associat-
ed with executive dysfunction and memory deficits in melancholic (as defined by
DSM-IV-TR criteria) depressed patients. They also reported no significant difference
between plasma and salivary cortisol levels in melancholic vs. none- melancholic
groups. The only significant correlation between cortisol values and CANTAB tests,
either mnemonic or prefrontal, used for cognitive function assessment was an asso-
ciation between morning salivary cortisol and the ID/ED total errors.

In accordance, Mitchell et al. (1990) and Joyce et al. (2002) compared a range of
basal measures in patients with melancholic depression defined by several criteria.
Mitch-ell et al. compared such diagnostic systems as ICD-9, DSM-III, RDC and
Newcastle Scale melancholic/endogenous phenotypes, while Joyce et al. (2002)
compared DSM-IV based diagnose of melancholia with that focusing on CORE
measures. In the second study, there was a significant correlation with basal corti-
sol: in male patients defined as melancholic by CORE criteria.

A few studies mentioned above about DST results also assessed the association
of basal cortisol levels with particular symptoms. Casper et al. (1987) reported that
basal plasma cortisol was significantly associated with weight and appetite loss;
Kaestner et al. (2005) indicated that high baseline cortisol levels correlated with
HAM-D severity and the presence of weight loss.

Basal ACTH
 There were three studies which addressed basal ACTH levels, of which none indi-
cated elevated plasma ACTH levels in melancholic patients.

Wong et al. (2000) who used both DSM-III-R and RDC criteria to define the me-
lancholic subtype failed to find a significant association in the melancho-
lic/endogenous group with elevated basal plasma ACTH; however, since they ob-
served elevated cortisol levels, they also reported that plasma cortisol-to-ACTH ratio
was significantly elevated in melancholic patients compared to controls. Similarly,
Joyce et al. (2002) failed to show elevated ACTH in melancholic patients when
applying either DSM-IV or CORE criteria. Finally, Gomez-Gil et al.(2010) got their
negative results when applying the NEDDI criteria.

Basal CRF

There were two studies which reported basal CRF levels. Wong et al. found that
melancholic patients did not differ from controls in their levels of basal CRF which
was disproportional about elevated basal cortisol. Similarly, Joyce et al. (2002) failed
to demonstrate elevated basal CRF using either DSM-IV or CORE definitions.

Atypical vs non-atypical or controls

We selected nine studies focusing on the atypical depressive subtype or its charac-
teristic features. Among this group, we indicated: studies focusing on DST (n=4),
studies assessing basal cortisol either in blood, urine, or the CSF (n=7), one study
assessing basal ACTH levels (n=1), two studies using desipramine stimulation (n=2),
a study using dextroamphetamine stimulation (n=1); a study using oCRH stimulation
(n=1).

Dexamethasone Suppression Test (DST)

Of the four studies assessing dexamethasone suppression rates in atypical patients,

only two used standard DSM-based criteria.
Levitan et al. (2002) evaluated DST response in 8 female patients with DSM-IV-
defined atypical MDD vs. 11 healthy controls and demonstrated that atypical patients
had higher rates of suppression vs. controls (91,9% suppression in atypical depres-
sive patients vs. 78,3% in controls). However, it is notable that the authors used a
lower dose of dexamethasone than usually administered: they used both 0.25 and
0.5 mg dosages, and significant results were reported with the latter (Levitan et al.,
2002).
Stewart et al. (2005) suggested stratifying patients into late/nonchronic atypical and
early/chronic atypical subtypes. Patients with early/chronic atypical had significantly
lower mean 3 h afternoon cortisol levels and 4:00 p.m. post- dexamethasone cortisol
levels than compared to late/nonchronic atypical (Stewart et al., 2005). This indicates
that the course of illness may also play an important role in the function of the HPA-
axis and that it may also contribute to the heterogeneity of atypical depression. How-
ever, since the study had no control group, it is difficult to draw conclusions as to
whether there is hypocortisolism in atypical patients compared to controls.
Casper et al., (1988) focused on somatic symptoms such as hypersomnia (n=23)
and overeating (hyperphagia, n=22), looking at these two symptoms separately, with
n=15 out of the 22 patients with hyperphagia also demonstrating weight gain, all
measured by SADS. The groups were compared with MDD patients who exhibited
neither of the symptoms and with matched controls. The study did not demonstrate
any increase in DST response in hypersomnia/overeating patients. However, it is an
important observation that patients with atypical features did not differ from controls.
In patients presenting with hypersomnia, there were significantly higher rates of nor-
mal suppression than in «non-atypical» patients, and the latter was similar to con-
trols.

Thase et al. (1989) compared a subgroup of bipolar depressed outpatients with

«anergic» depression which they defined using own operational criteria as manifes-
ting with «anergia» (score 2 on Hamilton scale item 13), «psychomotor retardation»
(score of 2 or more on item 8), and «reversed vegetative symptoms» where weight
gain was defined as an increase in weight of 2.2 kg or more, and hypersomnia as
increase of 1 hour or more compared to normal sleep duration). The authors focused
on identifying EEG disturbances and DST response in those patients compared to
controls. Only 3 (13%) of the patients were non-suppressors, even considering the
somewhat lower threshold for defining non-suppression (4mg/dl vs. the more com-
mon 5 mg/dl). Likewise, only 6 of the patients had baseline cortisol levels higher than
15 mg/dl. Authors also demonstrated that patients, and particularly 6 patients with
hypercortisolism, had decreased REM latency values.

Other challenge tests

Two of the studies assessed the levels of cortisol following a challenge test with 75
mg desipramine.

Asnis et al. (1995), compared a group of 17 patients diagnosed as suffering from

atypical depression to 55 patients not matching atypicality criteria. The criteria for
atypicality were similar to those of DSM-IV, except only one of symptoms additional
to mood reactivity (hypersomnia, hyperphagia, leaden paralysis, or rejection sensitiv-
ity) was obligatory for the diagnosis instead of two. It is striking how different the
phenotypes of patients included in this study could be compared to those mentioned
above. Patients were compared on their response to 75 mg of desipramine, which is
a challenge test for noradrenergic function. Although basal cortisol levels did not dif-
fer significantly between groups, post-DMI cortisol was significantly higher in atypical
group vs. non-atypical, which suggests that this group may have a less impaired
noradrenergic system compared to MDD patients without atypical features.

McGinn et al. (1996) studied patients from the same cohort as Asnis et al.(1995).
However, they stratified patients as having mood reactivity alone (n=29), having de-
pression with atypical features as defined by mood reactivity plus one of the four
atypical symptoms (n=33), and MDD not matching atypical criteria (n=52). The main
conclusion of the study was that AD patients, similar to the previous study, had a
significantly higher cortisol response to DMI.

Apart from DST, Stewart et al. used dextroamphetamine challenge in their sample.
The difference in post-dextroamphetamine cortisol levels did not reach statistical
significance although there was a trend for higher numbers in the early/chronic atypi-
cal group (Stewart et al. 2005).

Finally, Joseph-Vanderpool et al. (1991) examined HPA-axis function in patients with

seasonal depression characterised by atypical features such as reverse vegetative
symptoms. The authors used a challenge test with oCRH, which did not discriminate
between the subjects and controls. ACTH response to oCRH was delayed and re-
duced in Seasonal Affective Disorders patients.

Baseline measures

Baseline cortisol levels were assessed in seven studies in the group, while basal
ACTH was only mentioned in one study.

Both Asnis et al. (1995) and McGinn et al. (1996) who reported basal cortisol lev-
els in ADDS-assessed atypical depressive patients showed no difference be-tween
the patients and controls. Joseph-Vanderpool et al. (1991) reported a trend towards
lower basal cortisol in SAD patients vs. healthy controls which, however, was only
significant at 22.00.

Studies that applied DSM-IV criteria showed slightly differing results. Anisman et al.
(1999), compared 31 atypical MDD with 14 non-atypical MDD and 15 atypical dys-
thymic with 14 non-atypical dysthymic patients, assessing among other markers, on
basal cortisol and ACTH levels, and demonstrated significantly decreased basal cor-
tisol, but increased ACTH levels in atypical patients vs. controls.

Stewart et al. (2005) who also used DSM-IV criteria in their comparison of atypical
depression with various courses demonstrated that Patients with early/chronic atypi-
cal had significantly lower mean 3 h afternoon cortisol levels (Stewart et al, 2005).

Finally, Casper et al. (1988) and Levitan et al.(1997) used reversed vegetative symp-
toms as criteria for atypicality. The former showed no differences in any of the
measurements - i.e. plasma, CSF or urinary cortisol,and controls. At the same time,
Levitan et al. (1997) demonstrated a significant negative correlation between the
symptoms of hypersomnia and carbohydrate craving and basal cortisol values

Atypical depression vs melancholic depression

Studies directly comparing the function of the HPA-axis in melancholic vs. atypical
depression are rather scarce. Practically, our search only yielded seven relevant ar-
ticles. Of them, the majority (n=5) focused on baseline cortisol measures. There
were two studies that assessed basal ACTH as well, one study that assessed DST;
and one study that applied Dex/CRH test. The results of the studies are summarised
in Table 3 below.

Challenge tests

The correlation between personality disorders, depression subtypes, and DST

suppression rates was studied in 50 patients by Fountolakis et al.(2004). The au-
thors reported the results of DST in 14 atypical patients and 16 melancholic patients
defined by DSM-IV criteria. Other groups of patients included those with a «somatic
syndrome» as defined by ICD-10 (n=32, partly overlapping with atypical and melan-
cholic groups) and 9 patients without a clearly defined phenotype. The authors did
not observe any significant correlations between DST suppression rates with any of
the phenotypes; however, the largest proportion of non-suppression was observed in
the atypical group (42,85%), what goes against our previous described findings.
They also demonstrated the accumulation of cluster B personality disorders in the
atypical group, although that was not significant. The possible limitations of this study
are a small sample size and, possibly, the lack of healthy control group (Fountolakis
et al., 2004).

An elaborate study of stress reactivity patterns was performed by Heinzmann et

al. (2014). This study primarily focused on mice divided into three phenotypes based
on their stress response patterns; its second stage involved human partici-pants. Un-
like previous studies, the main criterion of patient grouping was not their depression
subtype, rather, patients were divided into high (hHR), intermediate (hIR) and low
(hLR) responders according to their cortisol response in the Dex/CRH test. Although
authors did not identify patients as having particular depressive subtypes, they ap-
plied HDRS subscale of non-atypical depression symptoms. Patients in the hLR
group showed less sleep disturbance, less appetite loss and less weight loss than
hHR patients. At the same time, hHR patients showed a strong trend towards higher
‘agitation’ scores and increased active stress-coping behaviour compared to hLR pa-
tients (Heinzmann et al. in 2014).

Basal cortisol and ACTH measures

The first study to compare the basal measures between the two subtypes was a
study by Elizabeth Young et al. (2001). Authors investigated the whether cortisol se-
cretion reflects a central CRF dysregulation or represents altered adrenal gland func-
tion. Over a period of 24 hours, ACTH and cortisol levels were assessed at 10-min
intervals in a sample of 25 premenopausal women and 25 healthy controls. Re-
garding depression subtyping, compared patients meeting RDC criteria for endoge-
nous depression (n=6) with patients meeting DSM-IV criteria for atypical depression.
They found no significant differences in either mean 24-hour plasma cortisol or uri-
nary cortisol secretion between groups, including between patients and controls, alt-
hough mean cortisol values tended to be higher in endogenous group vs. controls.
Regarding ACTH, the only significant findings were those regarding basal ACTH,
which was significantly increased in depressed patients in general vs. controls, and
so was the AUC for basal cortisol. No other significant differences were identified.
The obvious drawback of the study was essentially the small sample size which
comprised only 6 and 7 patients in phenotypes of interest.

Brouwer et al.(2005), recruited a bigger total sample (n=113) of MDD patients. How-
ever, the numbers of patients meeting DSM-IV criteria for atypical or melanchol-ic
depression were relatively low (32 and 25 patients). The authors analyzed an ar-ray
of endocrine measures including serum and urinary cortisol levels. In subtype analy-
sis, only serum cortisol was significantly lower in atypical depressed patients vs.
those not matching either subtype.

Karlovic et al. (2012) compared DSM-IV defined 23 melancholic depressed pa-

tients, 23 atypical depressed patients and 18 healthy controls on the levels of serum
cortisol (following a single morning blood test). The authors demonstrated a sig-
nificant difference between melancholic and atypical subtypes in the levels of morn-
ing cortisol, where the melancholic group showed an increased cortisol level while
the atypical group was not different from controls.

Cizza et al. (2012), recruited 89 female patients from the POWER (Premeno-
pausal, Osteoporosis, Women, Alendronate, Depression) having reported a depres-
sive episode in the past 3 years. According to DSM-IV criteria, 51 patients had mel-
ancholic depression, 16 presented with atypical features, and in 22, no subtype crite-
ria were met. The 24-hour sampling of plasma ACTH and cortisol yielded no signifi-
cant difference in 24-hour cortisol plasma cortisol between groups. However, plasma
ACTH was significantly higher in the atypical subtype vs. the control group (F (1, 83)
= 4.01, p<0.05) vs. controls. A group by time interaction demonstrated that ACTH
was elevated in the atypical group only in the daytime, with greatest differences ob-
served from 10 AM to 5 PM. Besides, after adjustment for total body fat, the mean
24-hour adjusted log leptin value was elevated in the melancholic subgroup, as
compared with controls (Cizza et al. 2012).
 The biggest patient sample analyzed so far was that recruited by Lamers et al.,
(2012), from the NESDA cohort. Authors compared some inflammatory, metabolic
markers, saliva cortisol awakening curves, and diurnal cortisol slope in 111 chronic
depressed patients with melancholic depression and 122 patients with atypical de-
pression. However, their labels did not refer to DSM classifiers, rather to results of a
latent class analysis performed using CIDI questionnaires, which showed no signifi-
cant effect of the measure of mood reactivity or interpersonal sensitivity, but a robust
effect of weight and sleep characteristics. In attempts to make the study more homo-
genous, the authors included only chronic severely depressed patients, as this ca-
tegory showed the most stable patterns of depressive symptoms. They demons-
trated that the atypical subtype differed from the melancholic subtype on a whole
range of symptoms, including area under the curve on the ground (AUCg) and diur-
nal cortsol slope measures (decreased in the atypical group) (Lamers et al., 2012).

Longitudinal studies

It is of note that due to a range of factors, and first of all the complexity of the
tests, there is a substantial lack of studies assessing the function of the HPA axis
across both phenotypes of interest longitudinally. Our search yielded only 5 articles
which used a longitudinal design. The first study found with the search terms was
that by R.G.Haskett et al. (2005), where the authors assessed the changes in the
DST response across several weeks in hospitalized patients. This methodology was
grounded in the previous observations that depressed patients who completed a
DST on day 2 of hospitalization had a higher frequency of cortisol nonsuppression
(71%) than depressed patients who were tested on days 3-6 (33%; Coccaro et al.,
1984). The study demonstrated that although there was an overall decrease in the
rates on non-suppression in both the patients with endogenous depression and con-
trol subjects (from 55 to 36%), this was mostly accounted for by the difference in the
control group, while the sensitivity of the test for the endogenous depressive patients
did not change significantly (Haskett et al., 1987).
In 1997, Steiger et al. performed a study assessing the levels of basal plasma cor-
tisol in 12 endogenous depressive patients diagnosed using RDC criteria on ad-
mission and post-treatment. They demonstrated that in ED patients, plasma cortisol
was significantly elevated on admission compared to healthy controls, and that it was
also significantly reduced after treatment, indicating that elevation was specific to
acute endogenous depression.
A study by Kaestner et al (2005) also employed a longitudinal design - pa-tients
were assessed on admission (t1) and after treatment, in remission (t2). The study
focused on assessing a range of factors, basal including basal cortisol and ba-sal
ACTH, in unmedicated, acutely depressed melancholic patients (n=37) compared
with 37 controls. They demonstrated that on admission, both cortisol and ACTH were
elevated in melancholic patients compared to controls, but not when compared with
non-melancholic patients. Plasma ACTH was still increased in melancholic patients
in remission compared to controls. At the same time, cortisol was elevated in acutely
depressed melancholic patients only, but not in remitted ones (Kaestner et al.,
2.005).
Pintor et al, 2013, measured ACTH and cortisol response using synthetic hu-
man CRF challenge in relapsing, non-relapsing and partially relapsing patients with
DSM- IV melancholic depression over a follow-up period of two years. In terms of the
overall comparison of depressed and healthy groups, significant differences were
observed for post-CRF ACTH, cortisol levels and for area under cortisol curve bet-
ween healthy controls and the three groups of melancholic patients.

Finally, the only article assessing the longitudinal course of HPA-axis abnormali-
ties in atypical depression is that by Geracioti et al., 1992, and represents a case
study of a female patient diagnosed with atypical depression across 6 months. The
results obtained by the authors correspond with the notion that atypical depression
has a different pattern of abnormalities: the patient was eucortisolemic in an acute
phase of depression, and further cortisol levels showed a negative correlation with a
deterioration of depressive symptoms.

Discussion

The key problem in diagnosis is the fact that elaborate classification systems that ex-
ist today are solely based on subjective descriptions of symptoms. Such detailed
phenomenology includes the description of multiple clinical subtypes; however, there
is no biological feature that distinguishes one subtype from another. Integrative ap-
proaches to understanding complex health issues can transcend disciplinary and
knowledge boundaries and provide opportunities to view phenomena from diverse
perspectives. A future diagnostic criteria system in which aetiology and pathophysi-
ology are essential in diagnostic decision-making would bring psychiatry closer to
other specialities of medicine (Juruena et al. 2007). Thus, the heterogeneity of clini-
cal conditions encompassed under the concept of major depression seems to be one
of the limiters of these advances. Therefore, the identification of distinct subtypes of
depression may allow advances in these areas by allowing the identification of more
homogeneous groups of patients, both in clinical aspects and in those related to the
aetiology and pathophysiology of the disorder presented. This depends on many fac-
tors like severity and type of depression, genotype, and history of exposure to stress,
temperament, and probably resilience (Mello et al., 2007)
The analysis of the articles focusing on the differences in the function of the HPA-
axis depending on depressive subtype has revealed a range of sufficient pitfalls in
research methodologies. In this sense, although the concept of a melancholic de-
pression subtype, equivalent to the concepts of endogenous or psychotic depres-
sion, has a long history of psychiatry and is well defined (Sullivan et al., 2002,
Baumeister & Parker, 2012). The atypical subtype, in turn, encompasses a hetero-
geneous group of patients and has only recently been introduced into the DSM-IV as
a specifier. Moreover, although the literature has given extensive support to the va-
lidity of depression with atypical features as distinct from melancholia and depres-
sion without atypical or melancholic features, there is still a certain degree of disa-
greement among researchers about which particular symptoms constitute this speci-
fier, and whether such factors as mood instability and interpersonal sensitivity have
the same weight in the dichotomy as biological reversed vegetative symptoms. Thus,
the lack of well-defined diagnostic criteria to characterize these subtypes of depres-
sion is reflected in the diversity of nomenclatures used in the literature to define
these subtypes.
In this sense, in this systematic review, we find a significant variation in the terms
used in the articles to define the melancholic and atypical subtypes. Besides the var-
iation in definition, a major complication is presented by the variation in approaches
to challenging tests (e.g. dosage, time of response measurement and threshold used
to defining non-suppression), and the variety of classifications used in order to define
«melancholic» or «atypical» subtypes complicate the task of arriving at a steady
conclusion.
Recently we have published a systematic review comparing the neuropsychologi-
cal performance of melancholic patients to non-melancholic depressive patients), in-
cluding atypical depressives, and healthy controls (Bosaipo et al. 2017). In this
study, the findings suggest that melancholic may have a distinct and impaired cogni-
tive performance compared to non-melancholic depressive patients on tasks involv-
ing verbal and visual memory, executive function, maintained attention and span, as
well as psychomotor speed, this last mainly when cognitive load is raised (Bosaipo et
al. 2017).
Besides, although the literature is increasingly demonstrating distinct differences in
clinical, biological, anatomical and response to treatment characteristics be-tween
these two subtypes of depression, this debate is still ongoing. In this sense, the HPA
axis play a vital role in the distinction between these subtypes, since stressful life
events play a major role in the pathogenesis and onset of depressive episodes
(Kendler et al., 2002). According to some authors, stress could lead to the onset of
the first depressive episode in genetically vulnerable individuals, making them even
more sensitive to stress in a fast forwarding fashion, compatible with the kindling hy-
pothesis by Post (1992). With this, the individual would need less stress to trigger
new crises, and it would become more vulnerable to the reprint of new depressive
episodes before different, sometimes milder, stressors (Post, 1992). Also, adverse
experiences in early life have been associated with significant increases in the risk of
developing depression in adulthood, particularly in response to additional stressors
(Tofoli et al. 2011; Cohen et al., 2001; Juruena, 2014). Thus, as the HPA axis is acti-
vated in response to stressors, changes in the functioning of this axis, at any level of
its components, and its regulations may play a pivotal etiological role in the onset of
depressive disorders (Holsboer, 2000; Tyrka et al., 2008).
Among the studies included in this systematic review that evaluated patients with
melancholic depression, most (n=17) studies focused on DST response. Of them,
the vast majority did indicate significantly elevated degrees of non-suppression in
melancholic patients.
Nevertheless, there have been two studies, which did not demonstrate this signifi-
cant association. The one Hubain et al. (1996), who performed DST in a large sam-
ple of 155 Newcastle Endogenous Depression Diagnostic Index (NEDDI) - de-fined
endogenous patients vs. a similar number of non-endogenous patients, in fact, did
initially show was a statistically significant difference in the dexamethasone suppres-
sion test response at 1600 h, but when the effects of age and severity of depression
were controlled, those differences disappeared. In a study by Berger et al.(1984), the
majority of comparison groups were other psychiatric patients, which somewhat
complicates drawing conclusions about the melancholic-nonmelancholic dichotomy,
however, this study showed the importance of biological symptoms such as weight
loss as a factor in non-suppression, confirming the notion that research may need to
focus more on the vegetative symptoms of subtypes of interest.
The inconsistence of results may partly be influenced by different dexamethasone
doses and suppression thresholds that were used. However, our review has demon-
strated that the most dramatic differences lie between studies that used different ap-
proaches to defining melancholic depression. So, when RDC was used as a defini-
tion scale, those having a diagnosis of endogenous depression showed higher non-
suppression rates than those with probable endogeneicity. Patients with DSM-
defined melancholic features tended to show higher non-suppression in melancholia,
too. It is also of note that strong support of elevated post-dex cortisol in melancholic
patients comes from the studies which either focused on particular symptoms which
are characteristic of the melancholic subtype or considered patients were meeting
more than one diagnostic scale criteria (e.g. both DSM-IV and RDC) or used their
operational criteria of endogeneicity. Another suggestion made by a few authors is
that depression characterised both by melancholic features and DST-non-
suppression is, in fact, a distinct form of depression. This inference stems from the
observations that melancholic patients who are non-suppressors exhibit higher basal
cortisol levels as well compared to melancholic patients normally responding to DST.
Also, this increase seems to be associated mainly with melancholic depression with
psychotic symptoms (Contreras et al., 2007).
Evidence of elevated basal cortisol and basal ACTH in melancholic patients is
much less consistent. Approximately half of the selected studies failed to demon-
strate differing levels of cortisol in melancholic patients compared to non-melancholic
ones or controls. This may be due to differences in methodology, or differences in
the diagnosis of melancholia (e.g. RDC definite endogenous criteria showed a more
consistent association than DSM-III melancholic criteria). Regarding ACTH, none of
the studies showed alteration in this measurement compared to controls.
However, when studies of melancholia focused on particular biological symptoms
such as weight loss and appetite loss/insomnia or used their operational criteria for
endogeneity mainly focusing on vital symptoms, they reported significant increases
in basal cortisol. This means that there may be a stronger association with biological
symptoms rather than subtypes as a whole.
When atypical studies were evaluated, importantly, the majority of those, re-
regardless of the outcome measures, did not show a difference between atypical pa-
tients and control subjects, although there was a significant difference between atyp-
ical and melancholic patients. There were indeed studies (Levitan et al., 2002; Anis-
man et al., 1999), which showed significantly decreased post-DST and basal cortisol
in atypical patients. However, it is important to consider that while Levitan et al.
(2002) compared atypical patients to healthy controls and their results indeed may
suggest hypoactive HPA-axis; Anisman et al. (1999) compared patients to non-
atypical depressed patients, which rather indicates the difference with another sub-
type.
The studies directly comparing the function of the HPA-axis between melancholic
and atypical patient groups are scarce and difficult for analysis since their methodol-
ogies vary largely. In particular, of the 7 studies comparing HPA axis functioning be-
tween patients with melancholic and atypical depression, only 2 used challenge tests
(Fountoulakis et al., 2004; Heizmann et al., 2014.). Among them, only the one by
Heinzmann et al.,(2014) despite a very different design from the rest of DST studies
observed in the whole review, showed a significant difference between sub-types,
There was no consensus in the studies assessing basal cortisol, while Brouwer et
al. (2005) and Lamers et al. (2012) did show decreased cortisol levels in atypical pa-
tients vs. controls and also underline the difference be-tween melancholic and atypi-
cal groups), others only indicated that they were not different between atypical and
controls. Notably, the design employed by Lamers et al. (2012) showed no signifi-
cant effect of the measure of mood reactivity or interpersonal sensitivity, but a robust
effect of weight and sleep characteristics. At the same time, the authors only recruit-
ed severely depressed atypical patients, which may also have contributed to the
strength of association.
This systematic review also considered a separate group including only longitudinal
studies. Although these studies are just a few and are also different in methodologi-
cal aspects, it could be suggested that in melancholic depression, elevated HPA-axis
function is a state rather than a trait characteristic, i.e. that remit-ted patients have
lower basal and post-challenge cortisol levels compared to acutely depressed pa-
tients. At the same time, the case study - and the only longitudinal study of the atypi-
cal subtype - showed a negative correlation between the severity of depressive
symptoms and cortisol levels, thus supporting the idea of the dichotomy.
Thus, although data in the literature seem to confirm that there are distinct pat-
terns of HPA axis functioning between the melancholic and atypical depression sub-
types, further studies with refining and homogeneous methodology are needed to
characterize this pattern better. These may be attributed to methodological differ-
ences, including varying challenges and doses and non-suppression thresholds, var-
ying availability of cortisol in urine, blood, saliva, or CSF. However, mainly the heter-
ogeneity of clinical conditions assessing the same endophenotype and incorporated
under the concept of major depression seems to be one of the limiters of these ad-
vances.
Novel advances in the methodology may shed light on the dichotomy in a more
precise manner. In particular, there are currently no published studies evaluating hair
and nail cortisol levels between the subtypes, but a few are underway. Be-sides,
speaking of challenge tests, among the articles included in this review, the majority
used the Dexamethasone Suppression Test. However, although the Dexa-
methasone Suppression Test remains widely used and widely studied as a biological
marker in psychiatry, this test has some limitations because of the pharmacokinetic
and pharmacodynamic characteristics of Dexamethasone that are very different from
cortisol. Unlike cortisol, Dexamethasone has low affinity to MR receptors. Therefore,
these studies allow us to investigate only the functioning of GR receptors in the sub-
types of depression (Pariante et al., 2002; Juruena et al. 2006). Future studies might
use different challenges, like MR antagonists, such as Spironolactone, MR agonists,
such as Fludrocortisone, the Prednisolone suppression test, which appears to bind
to MR as well as GR. MR function, and perhaps more important, MR/GR ratio, re-
mains understudied in depressed populations, and it seems like an interesting pro-
spect in this area (Juruena et al. 2013).
In general, this review has provided a rather convincing support for the presence
of a difference in HPA-axis activity between the two subtypes, melancholic and atyp-
ical depression, regardless of the classification. However, it is much more difficult to
conclude whether atypicality is associated with hypofunctional HPA-axis and en-
hanced negative feedback (such as implied in PTSD which was confirmed in a study
by Halbreich et al..1989), or simply is not different from controls. It may be that the
severity and the course of atypical depression, as well as the presence of particular
vegetative symptoms (hypersomnia, weight gain as opposed to interpersonal sensi-
tivity), are stronger predictors of decreased basal and post-challenge cortisol levels.
However, this is yet to be established in studies employing a more unified efficient
methodology.
The assessment of other factors potentially interfering with the dichotomy is out-
side the scope of this review. However, it should be noted that in the same studies
which showed conflicting results regarding hypocortisolism in atypical depression,
there was a much stronger association with elevated inflammatory factors (Lamers et
al., 2012). This has driven a novel appraisal of the two subtypes, suggesting that
while "typical"(or melancholic) depression has core pathophysiological features of
overactive HPA-axis, what we call "atypical" depression may rather be comprehend-
ed as immuno-metabolic depression (Penninx et al., 2016). The precise interaction
of potentially decreased activity of the HPA-axis with immune and metabolic abnor-
malities in the atypical subtype remains to be investigated.

CONCLUSION

The correct definition of depression subtypes remains a cornerstone in biological

research in affective disorders. The evaluation of study results is dramatically ham-
pered by the variation of definitions, and there is very little consistency between
research groups in what they name "endogenous" or "melancholic" depression. Our
review confirmed the presence of different HPA axis function between Melancholic
and Atypical Depression, and a trend towards a more robust association with biolo-
gical, or vegetative symptoms, or reverse vegetative symptoms, respectively. Pa-
tients with Melancholic depression are associated with increased cortisol levels, both
baseline and post different challenges. Moreover, the research data also suggest a
reduction of inhibitory feedback in patients with melancholic depression, demons-
trated by increased cortisol concentrations and the number of non-suppressive pa-
tients following HPA axis challenge, mainly dexamethasone. Whether the difference
between melancholic and atypical subtypes is better explained by the true hyperac-
tive HPA-axis in the latter or a rather normal function.
Future studies might need to focus on evaluating the symptom profiles in patients
with definite HPA-axis abnormalities to identify symptom constellations that are
strongly associated with neuroendocrine variations rather than rely on phenomeno-
logically defined subtypes. Moreover homogenize samples and methods, assessing
more naturalistic measures, like salivary, hair or nails cortisol levels. Further insights
into the dichotomy addressed in this review might be obtained from genetic and epi-
genetic studies of HPA-axis related genes in both subtypes, with an emphasis on the
presence of vegetative symptoms.
Disclosures

Dr MF Juruena is a Consultant Psychiatrist at University of Sao Paulo and Honorary

Consultant at South London and Maudsley NHS Foundation Trust (SLaM-NHS UK).
Professor AH Young is the Director of the Centre for Affective Disorders and is
supported by the National Institute for Health Research (NIHR); Biomedical
Research Centre (BRC) at SLaM-NHS UK and King’s College London. The views
expressed are those of the authors and not necessarily those of the NHS, the NIHR,
or the Department of Health. Dr MB and BA are a MSc student at Masters in Affec-
tive Disorders, KCL and has no conflicts of interest to declare. Only the authors were
involved in the study design and preparation of this report.

Role of funding source

Funding: This work was supported by Academy of Medical Sciences/Royal Society-
UK (MF Juruena), Only the authors were involved in the study design and prepara-
tion of this report.

Acknowledgements and Disclosurers:

This work was supported by Academy of Medical Sciences/Royal Society- UK; and
FAPESP (MF Juruena). Dr MF Juruena is a Consultant Psychiatrist at University of
Sao Paulo and Honorary Consultant at South London and Maudsley NHS Founda-
tion Trust (SLaM-NHS UK); Professor AH Young is the Director of the Centre for Af-
fective Diosrders and is supported by the National Institute for Health Research
(NIHR); Biomedical Research Centre (BRC) at SLaM-NHS UK and King’s College
London. The views expressed are those of the authors and not necessarily those of
the NHS, the NIHR, or the Department of Health. Dr MB and BA are a MSc student
at Masters in Affective Disorders, KCL and has no conflicts of interest to declare.

Conflicts of Interest:
MF Juruena has within the last year received honoraria for speaking from GSK,
Lundbeck and Pfizer. AH Young received honoraria for lectures and advisory boards
for all major pharmaceutical companies with drugs used in affective and related dis-
orders. Investigator-initiated studies from AZ, Eli Lilly and Lundbeck. Dr MB and BA
has no conflicts of interest to declare.
 REFERENCES

Amsterdam, J. D., Maislin, G., Gold, P., & Winokur, A. (1989). The assessment of abnormalities
in hormonal responsiveness at multiple levels of the hypothalamic- pituitary-adrenocortical
axis in depressive illness. Psychoneuroendocrinology, 14(1– 2), 43–62.
Anisman, H., Ravindran, a V, Griffiths, J., & Merali, Z. (1999). Endocrine and cytokine corre-
lates of major depression and dysthymia with typical or atypical features. Molecular
Psychiatry, 4(2), 182–8.
Asnis, G. M., McGinn, L. K., & Sanderson, W. C. (1995). Atypical depression: Clinical aspects
and noradrenergic function. American Journal of Psychiatry, 152(1), 31–36.
Banki, C. M., Arato, M., et al. (1986). Associations among dexamethasone non-suppression
and TRH-induced hormonal responses: increased specificity for melancholia. Psychoneuro-
endocrinology, 11(2), 205–211.
Barocka, A., Pichl, J., Beck, G., & Rupprecht, R. (1987). Factors interfering with the 1 mg
dexamethasone suppression test in depression. Pharmacopsychiatry, 20(6), 258–261.
Baumeister, H., Parker, G. (2012). Meta-review of depressive subtyping models. Journal of
Affective Disorders 139:126–140.
Benazzi, F. (2002). Can only reversed vegetative symptoms define atypical depression? Eur
Arch Psychiatry Clin Neurosci. Dec;252(6):288-93
Berger, M., Pirke, K. M., et al. (1984). The limited utility of the dexamethasone suppression test
for the diagnostic process in psychiatry. The British Journal of Psychiatry, 145, 372–382.
Bosaipo, NB, Foss, MP, Young, AH, Juruena, MF Neuropsychological Changes in Melancholic
and Atypical Depression: A Systematic Review Neuroscience & Biobehavioral Reviews, Feb
2017, 73:309-25
Brouwer, J. P., Appelhof, B. C., et al. (2005). Thyroid and adrenal axis in major depression: A
controlled study in outpatients. European Journal of Endocrinology, 152(2), 185–191.
Carroll, B.J., Feinberg, M., et al. (1981). A specific laboratory test for the diagnosis of melan-
cholia. Standartization, validation, and clinical utility. Arch Gen Psychiatry, Vol. 38.
Carroll, B. J., et al. (1982). The Dexamethasone Suppression Test for Melancholia.
Brit.J.Psychiat. (140) 292–305.
Casper, R. C., Swann, A. C., Stokes, P. E., Chang, S., Katz, M. M., & Garver, D. (1987).
Weight loss, cortisol levels, and dexamethasone suppression in major depressive disorder.
Acta Psychiatrica Scandinavica, 75(3), 243–250.
Casper, R. C., Kocsis, J., et al. (1988). Cortisol measures in primary major depressive disorder
with hypersomnia or appetite increase. Journal of Affective Disorders, 15(2), 131–140.
Cizza, G., Ronsaville, D. S., et al. (2012). Clinical subtypes of depression are associated with
specific metabolic parameters and circadian endocrine profiles in women: The power study.
PLoS ONE, 7(1).
Coccaro EF, Kavoussi RJ. (1994) Neuropsychopharmacologic challenge in biological
psychiatry. Clin Chem. 1994 Feb;40(2):319-27.
Cohen, P., Brown, J., and Smailes, E. (2001). Child abuse and neglect and the development of
mental disorders in the general population. Developmentand Psychopathology.981-99.
Contreras F, Menchon JM, Urretavizcaya M, Navarro MA, Vallejo J, Parker G. Hormonal differ-
ences between psychotic and non-psychotic melancholic depression. J Affect Disord.
2007:100:65-73.
Coryell, W. (2007), The facets of melancholia. Acta Psychiatrica Scandinavica, 115: 31–36.
De Kloet et al.(1998) Brain Corticosteroid Receptor Balance in Health and Disease. Endocrine
Reviews 19(3): 269–01
Diorio D, Viau V, Meaney MJ. The role of the medial prefrontal cortex (cingulate gyrus) in the
regulation of hypothalamic-pituitary-adrenal responses to stress. J Neurosci 1993; 13: 3839–
3847.
Evans, D.L., Nemeroff, C.B. (1987): The clinical use of the Dexamethasone Suppression test in
DSM-III affective disorders: Correlation with the severe depressive subtypes of melancholia
and psychosis. J Psychiat Res 21: 185
Fink, M. and Taylor, M. A. (2007), Resurrecting melancholia. Acta Psychiatrica Scandinavica,
115: 14–20.
Fountolakis, K.N., Iacovides, A. et al. (2004). Relationship among Dexamethasone Suppres-
sion Test, personality disorders and stressful life events in clinical subtypes of major depres-
sion: An exploratory study Annals of General Hospital Psychiatry. 3:15
Geracioti, T.D., Orth, D.N., et al. (1992). Serial cerebrospinal fluid corticotrophin-releasing
hormone concentrations in healthy and depressed humans. J Clin Endocrinol Metab 74:
1325–1330.
Gold, P.W. (2015). The organization of the stress system and its dysregulation in depressive
illness. Molecular Psychiatry; 20, 32–47; doi:10.1038/mp.2014.163.
Gómez-Gil, E., Navinés, R., et al. (2010) Hormonal responses to the 5-HT1A agonist buspi-
rone in remitted endogenous depressive patients after long-term imipramine treatment. Psy-
choneuroendocrinology. 35(4):481-9
Guechot, J., Lepine, J. P., et al. (1987). Simple laboratory test of neuroendocrine disturbance
in depression: 11 p.m. saliva cortisol. Neuropsychobiology, 18, 1–4.
Halbreich, U., Olympia, J., et al. (1989). Hypothalamo-pituitary-adrenal activity in endogenously
depressed post-traumatic stress disorder patients. Psychoneuroendocrinology, 14(5), 365–
370.
Haskett, R. F., Carroll, B. J., & Lohr, E. (1989). Comparison of Early and Delayed Inpatient
dexamethasone suppression tests.pdf
Heinzmann, J. M., Kloiber, S., et al. (2014). Mice selected for extremes in stress reactivity
reveal key endophenotypes of major depression: A translational approach. Psychoneuroen-
docrinology, 49(1), 229–243.
Holsboer, F. (2000). The corticosteroid receptor hypothesis of depression. Neuropsychophar-
macology. 23:477-501.
Hubain, P., Van Veeren, et al. (1996). Neuroendocrine and sleep variables in major depressed
inpatients: Role of severity. Psychiatry Research, 63(1), 83–92.
Joseph-Vanderpool, J. R., Rosenthal, N. E.,et al. (1991). Abnormal pituitary-adrenal responses
to corticotropin-releasing hormone in patients with seasonal affective disorder: clinical and
pathophysiological implications. Journal of Clinical Endocrinology & Metabolism, 72(6),
1382–1387.
Joyce, P.R., Mulder, R.T., et al. (2002). Melancholia: Definitions, Risk Factors, Personality,
Neuroendocrine Markers and Differential Antidepressant Response. Australian & New Zea-
land Journal of Psychiatry. Vol 36, Issue 3, pp. 376 - 383
Juruena MF, Cleare AJ, Papadopoulos AS, Poon L, Lightman S, Pariante CM. “Different re-
sponses to Dex and prednisolone in the same depressed patients”. Psychopharmacology
2006; 189 (2): 225-35.
Juruena, MF; Marques, AH; Mello, AF; Mello, MF. (2007). A paradigm for understanding and
treating psychiatric illness. Revista Brasileira de Psiquiatria, 29(Suppl. 1), s1-s2.
Juruena, MF, Pariante, CM; Papadopoulos, AS; Poon, L; Lightman, S, Cleare, AJ (2009).
Prednisolone suppression test in depression: a prospective study of the role of HPA axis
dysfunction in treatment resistance. British Journal of Psychiatry, 194, 342-49.
Juruena MF, Cleare AJ. Overlap between atypical depression, seasonal affective disorder and
chronic fatigue syndrome. Rev. Bras. Psiquiatr. 2007: 29(Suppl I):S19-26
Juruena, M. F., Cleare, A. J., Papadopoulos, A. S., Poon, L., Lightman, S., & Pariante, C. M.
(2010). The prednisolone suppression test in depression: Dose—response and changes
with antidepressant treatment. Psychoneuroendocrinology, 35(10), 1486–1491.
Juruena MF, Calil HM, Fleck MP, Del Porto JA. Melancholia in Latin American studies: a dis-
tinct mood disorder for the ICD-11. Rev Bras de Psiquiatr 2011:33(Suppl I):S48-58.
Juruena, MF. Early-life stress and HPA axis trigger recurrent adulthood depression Epilepsy &
Behavior, 38, 148-59, 2014.
Kaestner, F., Hettich, M., et al. (2005). Different activation patterns of proinflammatory cyto-
kines in melancholic and non-melancholic major depression are associated with HPA axis
activity. Journal of Affective Disorders, 87(2–3), 305–311.
Karlo i , D., Serretti, A., et al. (2012). Serum concentrations of CRP, IL-6, TNF-α and cortisol
in major depressive disorder with melancholic or atypical features. Psychiatry Research,
198(1), 74–80.
Kendler, K.S., Eaves, L.J., et al. (1996). The identification and validation of distinct depressive
syndromes in a population-based sample of female twins. Arch Gen Psychiatry; 53(5):391-9.
Kendler, K.S., Sheth, K., et al. (2002). Childhood parental loss and risk or first-onset of major
depression and alcohol dependence: the time-decay of risk and sex differences. Psychol
Med. 32(7):1187-94.
Klein, D.F., Davis, J.M. Diagnosis and drug treatment of psychiatric disorders. Williams &
Wilkins, Baltimore, 1969.
Klein DF. The pharmacological validation of psychiatric diagnosis. In: RobinsL, BarrettJ, eds.
Validity of psychiatric diagnosis. New York: Raven, 1989, 203–216.
Lamers, F., Vogelzangs, N., et al. (2012). Evidence for a differential role of HPA-axis function,
inflammation and metabolic syndrome in melancholic versus atypical depression. Molecular
Psychiatry, 18(6), 692–699.
Leventhal., A. M., Rehn, L.P. (2005). The empirical status of melancholia: Implications for
psychology. Clinical Psychology Review 25; 25 – 44
Levitan, R. D., Kaplan, A. S., et al. (1997). Low plasma cortisol in bulimia nervosa patients with
reversed neurovegetative symptoms of depression. Biological Psychiatry, 41(3), 366–368.
Levitan, R. D., Vaccarino, F. J., et al. (2002). Low-dose dexamethasone challenge in women
with atypical major depression: Pilot study. Journal of Psychiatry and Neuroscience, 27(1),
47–51.
Liu, Y., Yieh, L., et al. (2016). Metabolomic biosignature differentiates melancholic depressive
patients from healthy controls. BMC Genomics, 17, 669.
Maes, M., Schotte, C., et al. (1990). Clinical subtypes of unipolar depression: Part II. Quantita-
tive and qualitative clinical differences between the vital and nonvital depression groups.
Psychiatry Research, 34(1), 43–57.
Maes, M., Maes, L., et al. (1992). A clinical and biological validation of the DSM-III melancholia
diagnosis in men: results of pattern recognition methods. J Psychiatr Res. 26: 183-196.
Marques-deak, A. H., Neto, et al. (2007). Cytokine profiles in women with different subtypes of
major depressive disorder. Journal of Psychiatric Research. 41, 152–159.
McGinn, L., Asnis, G., & Rubinson, E. (1996). Biological and clinical validation of atypical
depression. Psychiatry Research, 60, 191–198.
Mello AF, Juruena MF, Pariante CM, Tyrka AR, Price LH, Carpenter LL, Porto JA. Depression
and stress: is there an endophenotype?. Rev Bras.Psiquiatr. 2007 29:s13-8.
Michopoulos, I., Zervas, I. M., et al. (2008) Neuropsychological and hypothalamic-pituitary-axis
function in female patients with melancholic and non-melancholic depression. European Ar-
chives of Psychiatry and Clinical Neuroscience, 258(4), 217–225.
Miller, K. B., & Nelson, J. C. (1987). Does the dexamethasone suppression test relate to sub-
types, factors, symptoms, or severity? Archives of General Psychiatry, 44(9), 769–74.
Mitchell, P., Smythe, G., et al. (1990). Hormonal responses to fenfluramine in depressive
subtypes. British Journal of Psychiatry, 157(OCT.), 551–557.
Mulder, R. T., Porter, R. J., & Joyce, P. R. (2003). The prolactin response to fenfluramine in
depression: Effects of melancholia and baseline cortisol. Journal of Psychopharmacology,
17(1), 97–102.
Nelson JC, Davis JM. (1997) DST studies in psychotic depression: a meta-analysis. Am J
Psychiatry; 154: 1497–1503.
O’Keane, V. and Dinan, T.G. (1991) Prolactin and cortisol responses to d-fenfluramine in major
depression: Evidence for diminished responsivity of central serotoninergic function.
Am.J.Psychiatry 148, 1009-1015.
Orsel, S., Karadag, H., et al. (2016).Diagnosis and Classification Subtyping of Depressive
Disorders: Comparison of Three Methods, Klinik Psikofarmakoloji Bülteni-Bulletin of Clinical
Psychopharmacology, 20:1,57-65
Orsel S, Karadag H, Turkcapar H & Kahilogullari A K (2010) Diagnosis and Classification
Subtyping of Depressive Disorders: Comparison of Three Methods, Klinik Psikofarmakoloji
Bülteni-Bulletin of Clinical Psychopharmacology, 20:1, 57-65
Parker, G., Roy, K., et al. (2002). Atypical depression: a reappraisal. Am J Psychiatry 159:
1470–1479.
Parker G, Paterson A, Hadzi-Pavlovic D. (2015). Cleaving depressive diseases from depres-
sive disorders and non-clinical states. Acta Psychiatr Scand. 131(6):426-33
Parker, G., McCrawa, S., et al. (2013). Validation of a new prototypic measure of melancholia.
Compr Psychiatry. 54(7):835-41.
Paslakis, G., Krumm, B.,et al. (2011). Discrimination between patients with melancholic de-
pression and healthy controls: Comparison between 24-h cortisol profiles, the DST and the
Dex/CRH test. Psychoneuroendocrinology, 36(5), 691–698.
Penninx, B. W., Milaneschi, Y., Lamers, F., & Vogelzangs, N. (2013). Understanding the somat-
ic consequences of depression: biological mechanisms and the role of depression symptom
profile. BMC Medicine, 11, 129.
Peselow, E.D., & Fieve, R.R. Depressive Attributional Style and the Dexamethasone Suppres-
sion Test: Relationship to the Endogenous/Melancholic Distinction and to Each Other. Psy-
chopathology 1992;25:173–182
Phelps EA, LeDoux JE. Contributions of the amygdala to emotion processing: from animal
models to human behavior. Neuron 2005; 48: 175–187.
Pintor L, Torres X, Navarro V, Martinez de Osaba MA, Matrai S, Gastó C. Corticotropin-
releasing factor test in melancholic patients in depressed state versus recovery: a compara-
tive study. Prog Neuropsychopharmacol Biol Psychiatry. 2007:31(5):1027-1033.
Porter, R., & Gallagher, P. (2006). Abnormalities of the HPA axis in affective disorders: Clinical
subtypes and potential treatments. Acta Neuropsychiatrica, 18 (5), 193-209.
Post RM. (1992). Transduction of psychosocial stress into the neurobiology of recurrent affec-
tive disorder. Am J Psychiatry. 149:999-1010.
Posternak, M.A., Zimmerman, M. (2001). Symptoms of atypical depression. Psychiatry
Res;104(2):175-81.
Quitkin, F.M., Stewart, J.W., et al. (1993) Columbia atypical depression: a subgroup of depres-
sives with better response to MAOI than to tricyclic antidepressants or placebo. Br J Psychi-
atry Suppl. 21:30–34.
Rubin, R. T., Poland, R. E., et al. (1987). Neuroendocrine aspects of primary endogenous
depression. Arch.Gen.Psychiatry, 44, 328–336.
Rush AJ, Weissenburger JE. Melancholic symptom features and DSM-IV. Am J Psychiatry
1994;151:489–498.
Schotte, C.K., Maes, M., et al. (1997). Cluster analytic validation of the DSM melancholic
depression. The threshold model: integration of quantitative and qualitative distinctions be-
tween unipolar depressive subtypes. Psychiatry Res. 71: 181-195.
Steiger, A., & Holsboer, F. (1997). Nocturnal secretion of prolactin and cortisol and the sleep
EEG in patients with major endogenous depression during an acute episode and after full
remission. Psychiatry Research, 72(2), 81–88. https://doi.org/10.1016/S0165-
1781(97)00097-8
Stetler, C., et al. (2013). Depression and Hypothalamic-Pituitary-Adrenal Activation: A Quantita-
tive Summary of Four Decades of Research. Psychosomatic Medicine 73:114–126
Stewart, J. W., Quitkin, F. M., et al. (2005). Defining the boundaries of atypical depression:
Evidence from the HPA axis supports course of illness distinctions. Journal of Affective Dis-
orders, 86(2–3), 161–167.
Stewart, J. W., McGrath, P. J., et al. (2007), Atypical depression: current status and relevance
to melancholia. Acta Psychiatrica Scandinavica, 115: 58–71.
Sullivan, P.F., Kessler, R.C. (1998). Latent class analysis of lifetime depressive symptoms in
the national comorbidity survey. Am J Psychiatry; 155(10):1398-406.
Sullivan, P.F., Prescott, C.A., Kendler, K.S. (2002). The subtypes of major depression in a twin
registry. J. Affect. Disord. 68: 273–284.
Thase, E., Ph, D., Jarrett, B., Kupfer, J., & Mallinger, G. (1989). E. Thase, (March), 329–333.
Tofoli, S.M.C., Baes, C.W., et al. (2011). Early life stress, HPA axis, and depression Psycholo-
gy & Neuroscience. 4, 2, 229 - 234
Tyrka, A.R., Wie,r L., et al. (2008). Childhood parental loss and adult hypothalamic-pituitary-
adrenal function. Biol Psychiatry. 63:1147-154.
Ulrich-Lai & Herman (2009). Neural regulation of endocrine and autonomic stress responses.
Nature Reviews Neuroscience 10, 397-409
Valdivieso, S., Duval, F., et al. (1996). Growth hormone response to clonidine and the cortisol
response to dexamethasone in depressive patients. Psychiatry Research, 60(1), 23– 32.
WHO(2017). Depression and Other Common Mental Disorders. Global Health Estimates. WHO
reference number: WHO/MSD/MER/2017.2
Winokur, G., Black, D. W., & Nasrallah, A. (1987). DST nonsuppressor status: Relationship to
specific aspects of the depressive syndrome. Biological Psychiatry, 22(3), 360–368.
Wong, M. L., Kling, M. A., et al. (2000). Pronounced and sustained central hypernoradrenergic
function in major depression with melancholic features: relation to hypercortisolism and cor-
ticotropin-releasing hormone. Proceedings of the National Academy of Sciences of the Unit-
ed States of America, 97(1), 325–30.
Yehuda R, Giller EL, Southwick SM, Lowy MT, Mason JW. (1991) Hypothalamic-pituitary-
adrenal dysfunction in posttraumatic stress disorder. Biol Psychiatry. 15;30(10):1031-48.
Young, E. A., Carlson, N. E., & Brown, M. B. (2001). Twenty-four-hour ACTH and cortisol
pulsatility in depressed women. Neuropsychopharmacology, 25(2), 267–276.

TABLE 1: Melancholic depressed studies

Thre
Dos shol
e d
Meas-
(For (for
Type of ure- Subtype
Au- chal- № of sub- Comparison
meas- ment Results
thor, leng chal- jects group
ure- time definition
year e leng
ment (s)
tests e
) tests
)

Dexa- Car- Dos 8 am 3, 4, 5, Endoge- n = 47 non- Sensitivity for

methas roll et e and 6 nous - endogenous: melancholic
one al., 1:1 4 pm μg/dL RDC n = 42
mg depression:
Sup- 1981a com- Endoge- 39% to 53%
(n= 11 pm (of them n =
pres- 183 pared nous - 32 Specificity:
sion ) Clinical
neurotic de- 85% to 97%
Test assess-
ment pression)
(DST) Dos Diagnostic
e 2: (match-
confidence:
2 ing RDC
mg in 98% 83% to 93%
(n= cases)
185
)
Banki 1 8 am 50 DSM-III n = 21 DSM-III 67% NS in
et al. mg ng/dL melancho- melancholic the Mel group
1986 3 pm MDD vs 26% in the
lic MDD
n = 15 schizophre-
healthy con- nia+adjustme
trols nt disorder
n = 20 schiz- group
ophrenia
n = 11 alco-
hol depend-
ence
n = 13 ad-
justment dis-
order

Evans 1 4 pm 5 μg/dl DSM-III n = 23 n = 23 non- Highest rate

et al., mg melancho- melancholic of non-
1986 11 pm suppression in
lic MDD MDD
psychotic pa-
n = 19 psy-
tients (95% vs
chotic MDD 78% in MEL
group at 5
μg/dl,
p<0.001);
Higher non-
suppression
rates in MEL
depression vs
non-MEL
(48%,
p<0.02)
Mil- 1 4 pm 7 μg/dl DSM-III+ n = 45 n = 39 MDD DST non-
ler et mg Endoge- MDD + suppression
11 pm + 5 psycho- correlated
al., nous - melancho
tic MDD with:
1987 RDC lia
+ 3 bipolar both melan-
Individual
depression cholic and
symp-
+ 3 schizoaf- endogenous
toms: subtype;
Yale De- fective
pression insomnia;
Inventory agitation;

loss of sexual
interest;

weight loss
Ru- 1 7 am 3,5 RDC «de- n = 40 n = 40 38% NS in
bin et mg μg/dl finite» healthy sub- the endoge-
3 pm endoge- nous group
al., jects
nous vs 12% in
1987 11 pm
controls

Cas- 1 8.30 6 μg/dl Loss of n = 38 n = 42 Depression

per et mg am appetite: MDD pa- with
items 12 MDD weight/appetit
al., tients wi-
4 pm on the patients e loss associ-
1987 thout
Hamilton with: ated with in-
10 pm scale (30), weight/appet creased basal
weight
32 on ite loss and post-Dex
loss
Vibes cortisol at all
and/or n = 80 con-
(31), time points
HSCL-90, appetite trol subjects compared to
item 19 loss
MDD without
(32), and weight/appetit
SADS-C e loss
item 228
(28).

Weight
loss: con-
tinuous
severity
measures
from 1lb
or more
Wino 1 8 am "Strong DSM-III n = 423 no healthy Melancholia
kur et mg suppres melancho- MDD controls significantly
al., and/o pres- associated
lic MDD patients
r4 sors": with high-
1987
pm cortisol degree
equal non-
to or suppression
lower
than (24% vs 9%,
1.5 p=0.01).
μg/dl,
n = 163
High
non-
suppres
pres-
sors:
(corti-
sol
equal
to or
greater
 than 6
μg/dl,
n=164)

Ber- 1,5 4 pm 5 mg/dl RDC/ICD/ SAMPLE SAMPLE 1: Control sub-

ger et mg 11 pm Newcastle 2: n = 75 jects: 12% NS
or n = 45 at 5 mg/dl
al., or Scale
MDD, of
healthy sub-
2,7% NS at 8
1984 1 them jects, of them mg/dl
8 mg/dl
mg n = 20 n = 24 DST 1 SAMPLE 2:
endoge- mg no significant
nous n = 51 DST differences in
SAMPLE suppression
1,5 mg
3: rates,
n = 93 however,
psychiat- SAMPLE 5: Higher NS
ric pa- n = 24 fasting rates among
tients, of patients patients with
them weight loss
n = 41 (p< 0.001)
endoge- SAMPLE 3:
nous 38,6% in ED
MDD, vs 7,4% in
n = 52 non-ED (5
other di- mg/dl)
agnoses 29,5% vs 8,7%
SAMPLE (5 mg/dl)
4: SAMPLE 4:
Higher NS
n = 93, of rates in neurot-
them ic depression
n = 19 vs ED at both
endoge- thresholds
nous
MDD
n = 74
patients
 with other
diagnoses

Baro 1 7 am 5 μg/dl ICD-10 n=26 n = 22 77% non-

cka mg 4 pm Endoge- neurotic suppression
et al., nous de- depression in endoge-
1987 pression +adjustme nous group
nt disorder vs 23% in
non-
endogenous

Am- 1 4 pm 5 μg/dl DSM-III n = 51 n = 21 non- Larger mean

ster- mg melancho- melancholic cortisol val-
dam lic MDD MDD ues for all
et al., n = 37 response val-
1989 ues in melan-
healthy con-
cholic non-
trols
suppressors
vs all other
groups and
healthy
controls
Halbr 1 4 pm not DSM-III n = 14 n = 21 No difference
eich mg speci- PTSD PTSD+M healthy con- between
et al., 11 pm fied PTSD patients
RDC DD-ED trols
with comor-
1989 MDD-
bid ED and
endoge- n = 23
controls; not a
nous de- MDD-ED single case of
pression non-
suppression in
the PTSD-ED
group.

Lower basal
cortisol and
higher dex
suppression
rates in
PTSD-ED vs
MDD-ED

Pasla 1,5 2 pm non- DSM-IV n = 26 n = 33 Higher post-

kis et mg speci- (SCID-IV) healthy con- Dex cortisol
al., 3 pm fied on MEL pa-
trols
2009 (as- tients (28.24
sessed ng/ml vs 12.1
as a ng/ml,
contin- p=0.02)
uous
meas-
ure)
Maes 1 8 am Non- Psycho- n = 96 for the vi- DST non-
M. et mg speci- pathologi- patients tal/nonvital supression
al., fied cal corre- assessed distinction: significantly
lates:
1989 for indi- n = 53 associated
(as- A. 14
sessed SCID vidual «nonvital» with symp-
as a symp- toms of ano-
items rexia, insom-
contin- toms/pres
uous B. ence of nia and early
meas- Clustering «vital» morning
ure) :
cluster; awakening;
1.«biologi
cal» clus- n = 33
ter: in- «vital»
crements
in FT, re-
sidual cor-
tisol, and
ACTH,
and by
decre-
ments in
basal
TSH, L-
TRP, and
L-TRP
ratio (all p
< 0.001)
2.«non-
biological
» cluster»
C. Vital (6
symp-
toms) vs
Nonvital
(7 symp-
toms)
syndrome
as validat-
ed in pre-
vious
studies by
Maes et
al., 1990
(Pt 1)
Pese- 1 8-9 5 μg/dl RDC de- 1) Meetin MDD mee- Morning post-
low mg am + finite en- g both ting neither DST plasma
E.D. 4-5 dogenous RDC and RDC nor cortisol; Af-
DSM- III ternoon post-
et al., pm subtype; DSM-III cri-
criteria DST plasma
1992 DSM III for mel- teria for en- cortisol and
melancho- ancholia: dogeneici-
lic sub- ty/melancholFrequency of
type n = 42 i a: abnormal
2) Meet- n = 43 DST signifi-
ing either cantly elevat-
RDC or Healthy sub- ed in «both»
DSM-III jects: vs «neither»
criteria subtype and
n - 29 in «both» vs
for mel-
ancholia: controls
n - 20

Hu- 1 4 pm 50 Newcastle n = 155 n = 155 Cortisol post-

bain mg ng/dl E ndoge- MDD MDD non- DST signifi-
et al., 11pm nous De- cantly elevat-
endoge- endogenous
pression ed in ED only
1996 nous
Diagnostic at 16.00
Index (p<0.01), but
(NEDDI): not at 23.00
>= 6 for Controlled for
endoge- age and sever-
nous <6 ity, post-DST
for non- 4 pm cortisol
endoge- no longer sig-
nous nificantly as-
sociated with
ED

Valdi 1 8 am ≥ 138 DSM-III n = 18 n = 29 non- Higher levels

vieso mg nmol/l. melancho- me- melancholic of DST non-
et al., 4 pm l suppression in
lic depres- lancholic MDD
1996 the MEL
11 pm sion MDD
n = 20 group
(p=0.004)
healthy vo-
lunteers
 Orsel, 1 8 am 3.5 g/dl DSM-III n = 38 DSM-III crite- Only non-
S. et mg melan- (DSM-III ria: reactivity
al., cholic de- criteria) differed
pression + n = 40 nonme- significant-
2010
cluster ly between
n = 40 lancholic
analysis non-
which «endoge- MDD,
suppressors
identified nous» and sup-
«endoge- cluster incl.
pressors
nous» and (included (p<0.01)
«non- n = 27
both
endoge- «simple» MDD
DSM mel
nous» and non-
subtype n = 4 BD1
mel
Discrimina patients) n = 2 Dysthy-
tors: mia
•early
morning n = 5 Depres-
awakening sive disorder
• distinct NOS
quality of
mood n = 2 Ad-
justment disor-
• feelings
der
of guilt,
non-
«Non-
reactivity
• suicidal endogenous»
ideation cluster: (in-
cluded 8 mel
•psychom patients) n = 38
otor
disorders

Other challenge tests

d Fenflu- O’Kean 30mg Baseline at Quanti- DSM 23 DMS 16 High
ramine e, 1991. 8:30; After tative, III III R healt baseline
challenge eve- compar- criteria criteria hy cortisol
ry 60 minutes, ison be- con- levels
for 5 hours tween trol were
groups sub- correlat-
jects ed with
severity
of
depres-
 sion
(p<0,01)
and
weight
loss
(p<0,01)
.
No dif-
ference
between
groups
after
chal-
lenge
d Fenflu- Mitch- 60mg 8 am (3 base- Quanti- Varied 16 DSM 14/ No sig-
ramine ell, 1990 line measures tative, accord III 12/ nificant
20 min, apart); compar- cord- 15 cortisol
after that chal- ison be- ing to 18 RDC /23 differ-
lenge and 5 tween used for ences
hourly groups crite- 15 ICD- each between
measures ria: 9 grou groups
DSM p
III 7 New-
RDC , castle
ICD-9, scale
New-
castle
scale

ACTH Amster- 250u 8:30 3 baseline Quanti- HDRS 1st: 16 1st: 1st: No
ster- g measures; after tative (9 with 11 differ-
dam, (ACT ACTH compar- melan- healt ence be-
1989 H) 30,60,90,120,1 ison cholic hy tween
80,240 min fea- con- groups
tures); trols at base-
line.
2nd: 72 Larger
(51 with increas-
melan- es in
nd
cholic 2 : cortisol
features, 37 in the
21 healt MDD
without) hy group
con- after
trols ACTH
(p<0,02)
;
2nd: No
 differ-
ence at
baseline;
no sta-
tistically
signifi-
cant dif-
ferences
between
groups,
but a
trend to
higher
cortisol
levels in
the mel-
ancholic
group
(P=0,31)
.
oCRH Amster- 1,0ug 8:30. 3 (every HDRS 26 11 No dif-
ster- /kg 15 minutes) (14 healt ference
dam, baseline melan- hy in corti-
1989 measures for cholic con- sol lev-
ACTH and cor- fea- trols els after
tisol levels; tures) chal-
after oCRH:0, lenge;
30,60,90,120,1 depres-
80, 240 min sive pa-
tients
had
lower
ACTH
response
com-
pared to
controls.
(p<0,05)
The ef-
fect was
larger in
melan-
cholic
features
(p<0,04)
Dex/CR Paslak- 100u 3 pm of day 2 Quanti- HDRS 26 33 Low
H test is, 2010 g (1 day after tative 21 moder- healt specific-
hCR DST) compar- ate to hy ity and
H ison severe con- low sen-
melan- trols sibility
cholic for the
depres- CRH/De
sion x test
(78.8%
and
30,8%
respec-
tively)

Basal Measures

Type of Authors, Time/source

Subtype defini- Control
mea- of mea- Sample size Results
year tion group
surement surement

Basal cor- Rubin et 26-hour corti- RDC «definite» n = 40 n = 40 DST NS vs

tisol sol curve: healthy Su-
endogenous
al., 1987 pressors:
Blood sam- subjects
levels - elevated
pling every 30
24-hour
minutes over cortisol
26 hours (11.4 vs
8.3
nmol/L,
urinary free p<0.01)
cortisol - elevated
nocturnal
nadir (3.1
vs 1.8
nmol/L, p<
0.04)
Only a
moderate
correlation
between
basal
serum and
UFC corti-
sol
Casper et Morning Loss of appetite: n = 38 n = 42 Basal
al., 1987 plasma corti- items 12 on the MDD patients MDD pa- plasma
sol obtained at Hamilton scale with: weight tients with- cortisol
8:30 a.m. on (30), 32 on Vibes loss and/or ap- out signifi-
days 9, 10, 12. (31), HSCL-90, petite loss weight/appe cantly as-
item 19 (32), and tite loss sociated
Evening
SADS-C item 228 with
plasma corti- n = 80 con-
(28). weight loss
sol level trol subjects
and ap-
drawn at 10
petite loss
p.m. on Day
Weight loss:
12.
continuous severi-
CSF cortisol ty measures from
sample - lum- 1lb or more
bar puncture
at 9 a.m. on
Day 11.
A 24-hour
urine predex-
amethasone
collection
assayed for
urinary free
cortisol (UFC)
completed at
10:35 p.m. on
Day 12.

Guechot et baseline sali- DSM-III DSM-III n = 40 «sec- Higher

al., 1987 vary cortisol Saint-Louis crite- Saint-Louis ondary de- saliva cor-
ria for primary criteria for pri- pressive» tisol in
11 pm saliva
endoge- mary en- endoge-
cortisol n = 20
nous/seconda ry doge- nous de-
«non-
depression nous/seconda ry pressives
depressive»
depression vs se-
condary
(p<0.05)/
non-
depressives
(p< 0.02)

Amsterdam basal morning DSM-III n = 26 n = 11 No diffe-

et al., 1989 plasma cor- healthy con- rence in
melancholic MDD (14 melancholic
tisol (8.30 am) trols basal corti-
features)
sol concen-
tration
between
melancho-
lic and
non-
melancho-
lic groups
Halbreich basal plasma DSM-III PTSD n = 14 n = 21 Lower
et al., cortisol RDC MDD-ED PTSD+MDD- healthy basal
ED cortisol in
1989 at 11 pm controls
PTSD-ED
n = 23 MDD-
vs MDD-
ED
ED

Wong et basal plasma DSM-III-R RDC n = 10 n = 14 Elevated

al., 2000 cortisol healthy basal cor-
controls tisol in
began 9.00-
MEL pa-
10.00 am
tients vs
every 30 min controls
for 30 hours (p<0.02)

Mi- basal morning DSM-IV (SCID- n = 20 n = 20 non- No signifi-

chopoulos cortisol levels IV) mel cant
elevation
et al., Salivary test n = 20
in MEL
2008 to assess cor- healthy
group
tisol levels controls
(three daily
samples:
morning,
08.00 a.m.
[CS1]; noon,
16.00 p.m.
[CS2] and
night,
23.00 p.m)

Paslakis 24-hour basal DSM-IV (SCID- n = 26 n = 33 Basal cor-

et al., plasma IV) healthy tisol
cortisol controls signifi-
2009 cantly ele-
vated in
MEL
group

Marquez- Basal plasma DSM-IV (SCID- n = 28 female n = 41 No signifi-

Deak et cortisol 8.00 IV) melancholic healthy cant
am MDD controls differences
al., 2007 n = 18 indicated
non- between
melancholic groups
MDD

Mulder et Basal plasma DSM-III-R n = 39 n = 69 non- No signifi-

al., 2003 cortisol at melancholic cant
melancholic
09.00 h, blood MDD, differences
MDD
drawn at 30- indicated
n = 20 con-
min intervals between
trol sub-
over 3 h 30 groups
jects
min
Joyce, Basal plasma DSM-IV me- n = 86 melan- n = 77 No diffe-
P.R. et al., cortisol lancholic MDD cholic patients
non-
rences on
any para-
2001 (13.00 - 15.00 n = 32 «severe- melancholic
meters
at 15-min ly melancholic» patients
vs between
intervals) patients
DSM-IV-
CORE defini- defined
CORE checklist tion: 116 groups
«broadly
CORE
defined»
definition:
melancholic
Basal cor-
patients;
tisol in-
39 «narrowly creased
defined» only in
melancholic male pa-
patients tients in
combined
broad+narr
ow me-
lancholic
vs non-
melancho-
lic groups
(p = 0.016)

O’Keane Basal plasma DSM-III-R, n = 23 n = 16 Elevated in

& Dinan, cortisol healthy endoge-
Newcastle scale
subjects nous pa-
1991 Single mea-
tients vs
sure at 8.30
controls
before
(t= 3,56;
d,l-
df=37,
fenfluramine
p=0.0001)
test
High base-
line corti-
sol corre-
lated with
HAM-D
severity
(r=0.97,
p<0.001)
and
weight loss
(r=0.85,
p<0.001).
Mitchell Baseline Four different def- n = 15 (ICD-9) n = 15 non- No signifi-
P. et al., plasma initions of endo- endogenous cant
n = 16 (DSM-
cortisol genicity/melanch (ICD-9) differences
1990 oly compared:
III)
in baseline
n = 14 non-
n = 18 (RDC) cortisol in
• ICD-9 endogenous
any
• DSM-III n=7 (DSM-III)
classifica-
• RDC (Newcastle
n = 12 non- tion
• Newcastle scale scale)
endogenous
(RDC)
n = 23 non-
endogenous
(Newcastle
scale)

Valdivies Baseline DSM-III me- n = 18 me- n = 29 non- No diffe-

o et al., plasma lancholic depres- lancholic MDD melancholic rences
cortisol at sion MDD between
1996 midnight depressed
n = 20
patients
healthy vo-
and con-
lunteers
trols

DSM-III melan-
Orsel, S. Basal plasma n = 38 (DSM- DSM-III Signifi-
cholic depres-
et al., cortisol 8 am III criteria) criteria: cantly ele-
sion+cluster anal-
vated basal
2010 ysis which n = 40 «endo- n = 40 non-
cortisol in
identi- genous» cluster melancholic
the
fied«endogenous» (included both MDD, incl.
«ëndoge-
and «non- DSM mel and
n = 27 nous» clus-
endogenous» non-mel pa-
«simple» ter
subtype tients)
MDD
Discriminators: n = 4 BD1
• early morning n = 2 Dys-
awakening thymia
• distinct quality of n = 5 De-
mood pressive
• feelings of guilt, disorder
non-reactivity NOS
• suicidal ideation n = 2 Ad-
• psychomotor justment
disorders disorder
«Non-
endogenous
» cluster:
(included 8
mel pa-
tients) n =
38
 Liu et al., cortisol as part The CORE scale n = 21 n = 58 an- Increased
2016 of 228 meta- for melancholic xious de- basal corti-
bolites depression; pression sol in me-
lancholia
Anxious depres- n = 100
vs healthy
sion defined as healthy
controls;
controls
number of comor-
bid anxiety disor-
ders on the
M.I.N.I. Interna-
tional Neuropsy-
chiatric Inter-
view > 0

Basal Wong et 24-hour basal DSM-III-R, n = 10 n = 14 No differ-

ACTH al. plasma ACTH healthy con- ence in
RDC (not speci- trols plasma
measures began 9.00-
10.00 am fied which used ACTH
every 30 min for diagnosis of
for 30 hours melancholia) between
patients
and con-
trols

Joyce, basal after- DSM-IV me- n = 86 melan- n = 77 No differ-

P.R. et noon plasma lancholic MDD cholic patients ences be-
non-
ACTH tween
al., 2001 n = 32 «severe- melancholic
groups
(13.00 - 15.00 ly melancholic» patients
vs independ-
at 15-min in- patients
ent of sub-
tervals)
CORE defini- type
CORE checklist tion: 116
«broadly
defined»
melancholic
patients;
39 «narrowly
defined»
melancholic
patients

Gomez- basal morning Newcastle Endog- n = 14 n = 15 No differ-

Gil et al. plasma ACTH enous Depression ence in
Diagnostic Index baseline
(NEDDI): ACTH
between
>= 7 for endoge- groups
nous
CRF Wong et Basal CSF DSM-III-R, n = 10 n = 14 No differ-
al., 2000 CRF - CSF healthy con- ence in
sampling be- RDC (not speci- trols baseline
gan at 09:00–
fied which used CRF be-
10:00 a.m.
and lasted for for diagnosis of tween
30 hours melancholia) groups

Joyce et Basal after- DSM-IV me- n = 86 melan- n = 77 No differ-

al., 2001 noon plasma lancholic MDD cholic patients ences in
non-
CRF basal
n = 32 «severe- melancholic
plasma
(13.00 - 15.00 ly melancholic» patients
vs CRF be-
at 15-min patients
tween
intervals)
CORE defini- groups
CORE checklist tion: 116 independ-
«broadly ent of sub-
defined» type
melancholic
patients;
39 «narrowly
defined»
melancholic
patients
Table 2 Atypical vs. non-atypical or controls
Dose Thresh
Type of Au- (For Meas- old (for № of Com-
measure- thor, chal- urement chal- Subtype def- sub- parison Results
ment year lenge time(s) lenge inition
jects group
tests) tests)

Dexame- Cas- 1 mg 8.30 am 5 μg/dl Hypersom- Sample n = 23 In pa-

thasone per et 4 pm som- 1: n = depressed tients
suppres- 10 pm 23 patients with hy-
al., nia/Increased
per-
sion 1988 appetite diag- MDD + without
somnia
test nosed with hyper- appe- only:
(DST) SADS somnia tite/weigh Signifi-
t increase cantly
or higher
hyper- levels of
somnia morning
Sample suppres-
2: n = sion vs
22 n = 22 non-
MDD + healthy atypical
incr.app subjects depressed
etite patients
(P < 0.04).
No differ-
ence be-
tween hy-
persomnia
patients
and con-
trols.
In pa-
tients
with
weight
gain only:
no signifi-
cant dif-
ference
between
patients
and con-
trols
In pa-
tients
with both
appe-
tite/weigh
t gain and
 hyper-
somnia
DST non-
suppres-
sion simi-
lar to con-
trols

Thas 1 mg 4 pm 5 μg/dl Own opera- n = 23 n = 26 13% DST

e et tional criteria «anergi healthy nonsup-
al., of anergia: c subjects pression
1989 1) definite bipolar» rate (3/26
anergia patients patients)
(score 2 on
HADRS-13)
2) psychomoto
r retardation
(score 2 or
more on
HADRS-13)
3) at least one
of two
associated re-
versed neu-
rovegetative
features
(weight gain
2.2 kg or more,
hypersomnia
as +1 hour of
extra sleep)
Levi- 0.25/0 8 am percent DSM-IV atyp- n = 8 n = 11 91,9%
tan .5 mg 3 am change ical depression female healthy suppres-
et al., scores patients subjects sion in
2002 atypical
depressive
patients vs
78,3% in
 controls;

Stew 1 mg 4 pm 5 μg/dl DSM-IV atyp- n = 84 no Lower

art et ical depression chronic healthy mean.
atypical controls
al., pa- post- dex-
2005 (compari- ame-
tients,
son be- thasone
of them tween
31 ear- atypical cortisol
ly- levels in
onset, groups)
ear-
53 late- ly/chronic
onset; atypical vs
61 late/nonch
chronic ronic atyp-
(dis-
ical pa-
thymia)
tients

Desipra As- 75 9 am: N/A ADDS n = 33 n = 81 Signifi-

mine nis et mg cortisol (meas- non- cantly
challenge al., levels ured as atypical higher
test every 15 a post-
1992 MDD
min for 1 conti- desipra-
hour; uous mine cor-
Imipra- varia-
mine vs tisol
ble)
placebo: (blunted
blood response)
test for 2 in atypi-
hours, cal group
(every 15 vs non-
min for atypical
cortisol
levels;
every 30
min for
desipra-
mine
levels)
 McG 75 9 am: N/A ADDS n = 17 Signifi-
n = 19
inn mg cortisol (meas- atypi- cantly
levels cal de- mood
et al., ured as blunted
every 15 pres- reactivi-
1995 a conti- response
min for 1 sives ty (MR)
uous to DMI in
hour; (AD) depres-
varia- AD com-
Imipra- sives,
mine vs ble) pared to
placebo: MR and
n = 36
blood controls
non-
test for 2
MR/AD
hours,
(every 15 MDD
minutes
for corti-
sol lev-
els; eve-
ry 30
min for
desipra-
mine
levels)
Dextroa Stew 0.15 Cortisol After DSM-IV n = 84 no HC No sig-
m- art et mg/k levesl 30 group nificant
phetamin al., g taken minutes differ-
e each 30 of Dex- ences be-
2005
stimulati min, tro
tween
on from infu-
groups
test 1pm to sion,
4pm; Af- cortisol (ear-
ter that levels ly/chroni
Dextro <1,5μg/ c atypical
stimula- dL vs
tion over were late/nonc
45 consid-
hronic
seconds, ered
and abnor- atypical)
blood mal
collected
every 15
min for
90min
oCRH Jose 100 Cortisol Quanti- DSM-III-R n = 10 n = 13 ACTH
stimula- ph- ng levels tative SAD healthy and corti-
tion test Van taken 15 be- pa- controls sol
minutes tween SAD with tients responses
der-
before groups reverse vege- to oCRH
pool, chal- tative symp- signifi-
J.R. lenge toms/ Major cantly
et al., (9am); depression blunted in
1991 then at with seasonal untreated
the time, pattern SAD
5, 10, 15, vs con-
30, 60, trols;
90 and
120 Delayed
minutes timing of
later the
ACTH
peak vs
controls

Table 2 Atypical vs non-atypical or controls. Basal cortisol

and ACTH levels

Time/sourc
Type of Authors, Control
e of Subtype Sample Results
measure- year measure- definition size group
ment
ment

Basal cor- Levitan at baseline DSM-III for n = 16 n = 14 Strong nega-

tisol al., 1997 plasma corti- Bulimia healthy con- tive correla-
«bulimia
sol Nervosa trols tion for hy-
at 8am nervosa»
persomnia
HDRS-29 to
with atyp- and basal cor-
assess re-
ical fea- tisol levels;
versed neu-
tures
rovegetative
for «carbohy-
symptoms
drate craving»
and basal cor-
tisol levels
Anisman baseline ADDS n = 31 n = 14 non- Decreased
et al., plasma cor- atypical atypical basal cortisol
DSM- depressed depressed
1999 tisol at 7am, in atypical
III/IV, patients, patients;
and each 10 depressed
minutes un- HAM-D-29 n = 15 n = 14 non- subjects vs
til 9:30am atypical atypical controls
dysthy- dysthymic
mic pa- patients
tients,
Casper et Basal plasma Hyper- Sample 1: n = 23 de- 1.Morning
al., 1988 cortisol 8 am somnia/ In- n = 23 pressed pa- plasma corti-
on days MDD + tients with- sol higher in
creased ap-
9,10,12 (fur- hyper- out appe- MDD without
ther petite diag- somnia tite/weight H or AI vs
averaged) nosed with increase or controls, no
SADS Sample 2: hypersomnia difference
Basal CSF
n = 22 otherwise
cortisol: 8-
MDD +
8.30 am
incr.appeti n = 22 2.Urinary free
Basal urinary te healthy sub- cortisol: high-
cortisol: jects er in all MDD
24-h speci- groups vs
men controls

3.No signifi-
cant differ-
ences in CSF
cortisol

Levitan et basal plas- DSM-IV n=8 n = 11 Lower basal

al., 2002 ma cortisol atypical de- cortisol in
(11 pm) pression subjects vs
controls

Stewart et basal plas- DSM-IV n = 84 no HC Lower mean

al., 2005 ma cortisol atypical de- group 3 h after-
pression noon cortisol
3-hour af-
ternoon cor- levels
tisol curve (N=21) in
early/chronic
atypical vs
late/nonchro
nic atypical
patients

Asnis et basal ADDS n=33 n=81 No differ-

al., 1992 morning ence be-
plasma cor- tween atypi-
tisol cal and non-
atypical de-
 pression (re-
ported as
“no effect
for group on
baseline cor-
tisol levels”)
McGinn basal ADDS n = 17 n = 19 MR No differ-
et al., morning patients ence in base-
plasma cor-
1995 n = 19 non- line cortisol
tisol between any
MR/AD-
groups
MDD pa-
tients
Joseph- basal DSM-III-R n = 10 n =13 Basal corti-
Vanderpo plasma sol levels NS
ol, J.R. et cortisol lower in pa-
al., 1991 (24-hour tients vs
curve) controls ex-
cept
at 22.00
when basal
cortisol in
patients was
significantly
lower vs
controls
(46.59 ±
27.59
nmol/L vs.
137.95 ±
71.73
nmol/L; p=
0.02)
Basal Anisman Basal ADDS n = 31 n = 14 non- Increased
ACTH et al., ACTH DSM- MDD atyp MDD; basal ACTH
1999 at 7am, and III/IV, in atypical
each 10 HAM-D-29 n = 15 n = 14 non- depressed
minutes un- dysthy- atyp subjects vs
mia dysthymia controls
til 9:30am

TABLE 3 ATYPICAL DEPRESSION VS. MELANCHOLIC DEPRESSION. BASAL

CORTISOL. BASAL ACTH, DST, DEX/CRH
 Definition
of
Type Definition
melanchol
of Authors, Measureme of atypical Comparis
ic Results
measureme year nt time subtype/ on group
subtype/
nt sample size
sample
size
Young et 24-hour RDC DSM-IV n = 25 nei- A trend
al., 2001 plasma cor- definite criteria; ther sub- towards
tisol: 9 am-9 endogenou n=7 type MDD elevated
am; s; cortisol in
definite
n=6
10-min in- endogenou
tervals s patients;
normal
24-hour uri-
cortisol in
nary cortisol atypical
patients.
No
significant
differences
.

Brouwer Morning DSM-IV DSM-IV n = 56 Serum

et al., MDD nei- cortisol
basal serum n = 32 n = 25
2005 ther sub- lower in
Basal cortisol type AD vs nei-
(before ther sub-
cortisol n = 113
10am) type MDD
control
24-h urinary only
subjects
cortisol

Karlovic Baseline DSM-IV DSM-IV n = 18 Serum

et al., morning se- cortisol
n = 32 n = 23
2012 rum cortisol signifi-
between cantly
8am and higher in
9am MDD-M
vs MDD-
A, and in
MD vs
control

Similar in
AD and
controls.
 Definition
of
Type Definition
melanchol
of Authors, Measureme of atypical Comparis
ic Results
measureme year nt time subtype/ on group
subtype/
nt sample size
sample
size
Lamers et Saliva corti- CIDI + CIDI + pre- n = 543 Lower
al., 2012 sol awaken- previous vious latent healthy AUCg,
ing curves latent class class analy- controls lower di-
(CAR 30, 45 analysis sis (own urnal cor-
and 60 (own oper- operational tisol slope
minutes lat- ational cri- criteria) in atypical
er. Addi- teria) of of patients vs melan-
tional sam- patients with persis- cholic pa-
ple at 11am. with per- tent chronic tients and
Diurnal cor- sistent depression vs controls
tisol slope chronic n = 111
depression
n = 122

Cizza et 24-hour se- DSM-IV DSM-IV n = 22 nei- No varia-

al., 2012 rum cortisol, n = 53 ther sub- tion in se-
n = 16
hourly, start- type MDD rum corti-
ing at 8am. sol levels
n = 44
healthy
controls

Young et 24-hour RDC DSM-IV n = 25 nei- No differ-

al., 2001 plasma n=6 n=7 ther sub- ence be-
ACTH type MDD tween
groups;
9 am-9 am;
10-min in- a trend
tervals toward
increased
Basal ACTH in
endoge-
ACTH
nous pa-
tients vs
controls.
Cizza et 24-hour se- DSM-IV DSM-IV n = 22 nei- Significant
al., 2012 rum cortisol n = 53 ther sub- increase in
n = 16
hourly, start- type MDD serum
ing at 8am. n = 44 ACTH in
atypical
healthy
 Definition
of
Type Definition
melanchol
of Authors, Measureme of atypical Comparis
ic Results
measureme year nt time subtype/ on group
subtype/
nt sample size
sample
size
controls group vs
controls

Fountolak 1 mg 4 pm 5 μg/dl DSM- DSM-

is et al., 11 pm IV/ICD-10 IV/ICD-10
DST
2004
n = 16 n = 14

DST/CRH Heinzman Dex: 0.05 11.30 pm AUC Based on n/a

n et al., mg/kg - 2 measureme AUC
2014 nt response to
mg/kg CRH
dose: 0.15 Dex/CRH,
patients
mg/kg.
(n=657)
were
divided
into:

hHR: 219

hLR:219

hIR: 219

TABLE 4 LONGITUDINAL STUDIES

Type of
Subtype
Au- Subtype Comparison
measurement
thors, Design + Results
(incl.dose
year definition group
sample size
and timing
 specifications)

Haskett Dexame- DST Endoge- RDC "definite Non- Non-

et al., thasone Sup- nous endogenous suppression
pression Test performed endogenous"
1987 MDD pa- rates signifi-
on Depression
(1 mg; depression tients, cantly lower at
admission incl."probabl second DST
4 pm-11 pm, e endoge- overall and in
(day 1-2)
5 ng/dL) and after nous" nonpendoge-
nous patients,
7-12 days
Non-
suppression
rates did not
significantly
decrease in the
ED group

Geraci- Basal plasma 6-month Atypical Clinical n/a Eucortisolemic

oti et cortisol follow-up in acute de-
al., Depression assessment pression (on
(biweekly admission)
1992 (the presence
tests at 7.30 of reversed Negative corre-
am) neurovegeta- lation between
tive signs) mood ratings
and serum cor-
tisol (r = - 0.36,
p = 0.002)

Steiger Basal plasma Time 1. On Endoge- RDC n = 25 nor- Basal cortisol

et al., cortisol admission nous De- mal controls significantly
1997 pression elevated in ED
(from 23.00 Time 2. vs controls;
Post-
until 07.00) treatment in ED on admis-
sion vs ED post-
treatment

Pintor Plasma Patients melan- DSM-IV n = 23 No differences

et al. ACTH levels followed- cholic pa- melancholic healthy between re-
2013 following up for 2 tients MDD, con- subjects lapsed, non-
100 μg hCRF years af- firmed by relapsed or
ter hCRF n = 62 partially re-
stimulation MES,
stimula- lapsed groups;
test NEDDI
tion Significant
 Plasma corti- differences
sol levels fol- NAUCC between pa-
lowing100 stratified tient groups
μg hCRF at three and controls
levels — on both
stimulation
<150; ACTH and
test
150 – 350 cortisol
and significant
>350 difference on-
μg/ml/mi ly between
n complete re-
lapse groups
and controls
Kaestne • Basal Time 1. On Endogenous DSM-IV + n = 16 non- Plasma ACTH
r et al., admission Newcastle melanchol- increased in
plasma
2005 Depression Endogenicity ic both acute
ACTH (8 Time 2. patients and remitted
am) Scale
Post- n = 21 MEL patients
6 points or n = 37
• Basal treatment more healthy
vs controls.
plasma controls Elevated
cortisol (8 plasma corti-
am) sol in acute
melancholic
vs control
groups only
Non-MEL pa-
tients not dif-
ferent from
control on ei-
ther measure

HIGHLIGHTS
 Different depressive subtype classification criteria may influence severity and
treatment.
 Depressive subtype is an important factor influencing Hypothalamus-Pituitary-
Adrenal (HPA) axis activity
 Melancholic patients has increased post-challenge cortisol levels than Atypical
Depressive patients
 Studies focusing on reversed vegetative symptoms, demonstrated a decrease in the
activity of the HPA axis in atypical depressives compared to controls, but the ma-
jority did not distinguish it from healthy controls.
 The correct definition of depression subtypes remains a cornerstone in biological
research in affective disorders.
 Future studies should consider epigenetic studies of HPA-axis related to both sub-
types, with an emphasis on vegetative symptom and standardized methodologies.