HPA axis and sleep: Identifying subtypes of major depression

Abstract

It is increasingly acknowledged that the diagnosis of major depression encompasses patients who do not necessarily share the same disease biology. Though the diagnostic criteria allow the specification of different subtypes, e.g. melancholic and atypical features, a consensus still has to be reached with regard to the clinical symptoms that clearly delineate these subtypes. Beside clinical characteristics, biological markers may help to further improve identification of biologically distinct endophenotypes and, ultimately, to devise more specific treatment strategies.

Alterations of the hypothalamus–pituitary–adrenal (HPA) axis and sleep architecture are not only commonly observed in patients with major depression, but the nature and extent of these alterations can help to identify distinct subtypes. Thus, a HPA overdrive, due to enhanced secretion of corticotropin-releasing hormone (CRH) and an impaired negative feedback via glucocorticoid receptors, seems to be most consistently observed in patients with melancholic features. These patients also show the clearest sleep-electroencephalogram (EEG) alterations, including disrupted sleep, low amounts of slow wave sleep (SWS), a short rapid eye movement (REM) latency and a high REM density.

In contrast, patients with atypical features are characterized by reduced activity of the HPA axis and ascending noradrenergic neurons in the locus coeruleus. Though sleep-EEG alterations have been less thoroughly examined in these patients, there are data to suggest that SWS is not reduced and that REM sleep parameters are not consistently altered.

While the atypical and melancholic subtypes of major depression may represent the extremes of a spectrum, the distinct clinical features provide an opportunity to further explore biological markers, as well as environmental factors, contributing to these clinical phenotypes.

Moreover, dysregulations of the HPA axis and sleep-EEG alterations can also be induced in rodents, thereby allowing alignment of critical biological aspects of a human disease subtype with an animal model. Such “Translational Research” efforts should help to develop targeted therapies for distinct patient populations.

• Major
• depression
• HPA
• sleep-EEG
• melancholia
• atypical
• age

Introduction

Today, the psychiatric diagnostic classifications used in the USA and Europe (the Diagnostic and Statistical Manual, DSM-IV-TR, and the International Classification of Diseases, ICD-10) offer clinicians a plethora of diagnoses. Most of the possible diagnoses, particularly for affective disorders, lack a distinct clinical description, and include no biological features to unambiguously delineate one diagnostic entity from another. This seems a particular dilemma for affective, including anxiety, disorders, as there is growing evidence for the existence of subtypes that show clinical differences and distinct biological features (Gold and Chrousos 2002; de Kloet et al. 2005; Ising et al. 2005). Not surprisingly, some critics of the current diagnostic conundrum have described major depression as a “pseudo-category, effectively homogenizing multiple expressions of depression” (Parker 2004).

Although today's classifications allow further specification of the clinical features of major depressive disorders, for instance distinction of melancholic, atypical and psychotic features, the criteria for the former two remain a matter of significant debate (Parker 2000; Parker et al. 2002, Parker et al. 2005; Parker 2005). Though the criteria for psychotic depression are less controversial, biological data question its validity as a biological subtype, as some of the changes noted in the sleep-electroencephalogram (EEG) and hypothalamus–pituitary–adrenal (HPA) axis function seem more a matter of severity of depression and age rather than psychotic feature per se (Kupfer et al. 1986; Thase et al. 1986; Stefos et al. 1998).

Some anticipate that a revision of the current DSM-IV-TR classification (the fifth version) will entail a more detailed clinical description of patients, including such features as the age of onset of the disorder, its course, the context in which it occurs, the familial background as well as both psychiatric and medical co-morbid disorders. Also, there is hope that the new classification will consider neurobiological features. This revision would help to determine the boundaries of specific disorders, and would thereby foster the identification of clinically relevant biological markers for distinct diagnostic entities in psychiatry (Kupfer 2005). However, availability of DSM-V is not expected before 2011, quite possibly even later.

In the meantime, it seems pertinent to endorse the search for clinical and biological features that help to delineate distinct affective disorders, in terms of their biology and clinical course. Sleep-EEG and stress hormone alterations, were the first biological changes reported in major depression several decades ago (Carroll et al. 1976b; Kupfer 1976). Today, there is a wealth of data describing a number of biological alterations, including sleep-EEG and hormone changes, in major depression. However, far fewer studies have used such measurements towards a biological differentiation within the heterogeneous population of depressed patients. Therefore, the present work focuses on studies that assessed the HPA axis and/or the sleep-EEG in association with key clinical characteristics to delineate rather than amalgamate distinct subtypes of major depression.

Clinical features to subtype major depressive disorder

The DSM-IV TR criteria for a diagnosis of a major depressive episode require that symptoms represent a change and are present during the same two-week period. At least one of the symptoms has to be either depressed mood nearly every day or loss of interest/pleasure in all activities. In addition, at least four of the following symptoms have to present as well: (i) significant weight loss (not from dieting) or weight gain, or decrease or increase in appetite, (ii) insomnia or hypersomnia, (iii) psychomotor agitation or retardation, (iv) fatigue or loss of energy, (v) feelings of worthlessness or excessive or inappropriate guilt, (vi) diminished ability to think or concentrate, or indecisiveness, and (vii) recurrent thoughts of death or recurrent suicidal ideation.

Once the criteria for a major depressive episode are met, the DSM-IV offers the possibility of four further specifiers: postpartum, melancholic, atypical or catatonic depression.

Melancholic features, as defined in the current revised edition of DSM-IV, include as a mandatory criterion loss of pleasure (anhedonia) and at least three of the following: depressed mood that is distinct, mood that is worse in the morning, early morning awakening, substantial weight loss, extreme feelings of guilt and psychomotor disturbances. These features have been reported to be more often observed in elderly than younger patients (Maes 2002; Parker et al. 2003; Brodaty et al. 2005). Besides age, studies also suggested that gender influences the clinical symptoms, with men presenting more often with melancholic features than women (Khan et al. 2006). This is in line with epidemiological data suggesting that the known gender bias, namely a much higher prevalence of depression in women than men, is restricted to atypical depression (Angst et al. 2002). With regard to psychotic depression, some consider this subtype merely a severe form of melancholic depression, implying that patients with psychotic features also show melancholic symptoms. The few studies exploring sleep-EEG and/or concomitant HPA axis measurements in this patient population seem to confirm this notion, in as much as patients with psychotic depression show the highest rate of non-suppression in the dexamethasone suppression test (DST) (Schatzberg et al. 1984, 1985) and pronounced sleep alterations (Kupfer et al. 1986; Thase et al. 1986). However, these studies also showed that a biological distinction between psychotic and non-psychotic depression is incomplete at best, as sleep-endocrine alterations distinguish less between absence or presence of psychotic features, but rather between younger and older patients and less and more severely depressed patients (Kupfer et al. 1986; Stefos et al. 1998). Also, recent data could not confirm a specific benefit of anti-glucocorticoid treatment in psychotic depression. In view of the paucity of data supporting psychotic depression as a separate biological entity, the ensuing discussion will focus on melancholic depression, with the hypothesis that psychotic features are in a further subgroup within the population of melancholia.

In contrast to melancholic and psychotic features, the atypical features specifier is an almost opposite clinical presentation, with mood reagibility as a mandatory, albeit controversial, criterion, and at least two of the following: (i) significant weight gain or increase in appetite, (ii) hypersomnia (10 h total or 2 h beyond normal), (iii) leaden paralysis and (iv) long-standing pattern of interpersonal rejection sensitivity.

While some authors questioned the usefulness of the last criterion, the reversed neurovegetative signs, hypersomnia and increased weight and appetite, may be more robust criteria to define a biologically distinct group of patients. Furthermore, Thase et al. (1995b) raised the hypothesis some years ago that reversed neurovegetative features were actually not atypical at all, but quite typical features of depression in young women. A recent elegant study seems to confirm this hypothesis: in opposite-sex twin pairs with major depression, women were more likely to present with increased weight, appetite, hypersomnia and fatigue, while men had a higher odds ratio to show insomnia and agitation during a major depressive episode (Khan et al. 2002). Further support for a subtype-specific gender bias in major depression is provided by an extraordinary epidemiological study from Zurich (Angst et al. 2002): the authors found that women were markedly overrepresented only among patients with atypical features. In addition to the gender bias, atypical depression has been associated with an earlier onset, a more chronic course and poor response to tricyclic antidepressants, pointing to a distinct biology (Quitkin et al. 1988, 1989, 1993; McGrath et al. 1992; Stewart et al. 2003, 2005)

Although the abandonment of an etiological in favor of a phenomenological concept in the DSM-III in 1980 makes a full alignment difficult between the patient groups diagnosed before and after the revision made in DSM-III, endogenous depression, recurrent depression with vegetative features and melancholic depression are likely to describe a similar group of patients. On the other hand, it can be assumed that a substantial number of patients formerly diagnosed with neurotic depression meet the criteria of atypical depression today.

Biological features

Key methodological approaches to detect alterations of the hypothalamus–pituitary–adrenal axis in major depression

Several methods have been explored to measure HPA axis activity in depression, following the initial observation that some patients show elevated plasma cortisol levels and reduced negative feedback inhibition (Asnis et al. 1981b). In general, the methods can be divided into: (i) more or less sophisticated measurements of basal plasma cortisol (and related hormones, such as ACTH) and (ii) challenge tests.

The former include basic measures such as the amount of free cortisol excreted over 24 h in urine, with no time resolution, and more sophisticated approaches, examining cortisol and ACTH in plasma samples collected frequently over 24 h or at least during the night, when cortisol reaches its nadir (Antonijevic et al. 2000b; Carroll et al. 2007). The two most widely used pharmacological challenge tests include the dexamethasone suppression test (DST), which explores negative feedback inhibition of the HPA axis after administration of dexamethasone (Dex), a synthetic glucocorticoid receptor agonist (Carroll et al. 1976a), and a more sensitive but less specific test that combines dexamethasone with subsequent administration of a bolus injection of the corticotropin-releasing hormone (Dex–CRH test) to also assess pituitary responsiveness (Heuser et al. 1994). For the DST, a cut-off value has been calculated, which allows clear distinction between subjects with and without suppressed cortisol values (suppressors vs. non-suppressors). Non-suppression has been reported in 40–55% of patients with endogenous or melancholic depression (Rush and Weissenburger 1994; Coryell 2007). In contrast, an exaggerated ACTH and cortisol secretory response in the Dex–CRH test is seen in a majority of patients with major depression, but lacks specificity for a particular subgroup (Heuser et al. 1994).

Further challenge tests include stress tests, such as the Trier-Social-Stress-Test (TSST), that measures cortisol and ACTH secretion in the context of a stressful situation (Kirschbaum et al. 1993). An exaggerated cortisol and ACTH response in this test has been linked to adverse early life events (Heim et al. 2000, 2004).

Key methodological approaches to detect alterations of the sleep-EEG in major depression

Sleep-EEG analysis, due to its complexity and detailed time resolution, is correctly considered a window on the sleeping brain (Dijk 1995). As sleep disturbances are among the most common symptoms of major depression and rank high among the residual symptoms (Fava 2006), the sleep-EEG offers an excellent tool to objectively assess sleep alterations and relate these to the pathophysiology of major depression.

Sleep-EEG assessments can be divided into (i) conventional and (ii) spectral sleep-EEG analyses. The former measure the time spent in different sleep stages (namely in REM sleep and non-REM sleep stages 1–4) as well as how long it takes to reach a specific sleep stage.

Spectral analysis provides detailed information with a high time resolution on the nature and frequency of the EEG waves in and across the different sleep stages. Thus, REM sleep is characterized by low amplitude, high frequency EEG waves similar to the awake EEG, but with REMs. Non-REM sleep typically comprises stage 2–4 sleep. Stage 2 sleep is characterized by sleep spindles, which are low amplitude, high frequency (in the range of 10–15 Hz) waves, spanning approximately one second. Stage 3 and 4 sleep, which, when taken together, are termed SWS, are characterized by the highest abundance of high amplitude, low frequency (0.5–4.5 Hz) delta waves. As sleep-EEG patterns follow a well-known time course, driven by both homeostatic and circadian influences, changes in the distribution of EEG activity during sleep can help to uncover distinct pathophysiological changes (Pace-Schott and Hobson 2002).

Methodological approaches to detect sleep-endocrine alterations in major depression

Since cortisol concentrations show a circadian rhythm, with trough levels during the early hours of the night and peak levels shortly after final awakening, frequent nocturnal cortisol measurements allow detection of both circadian alterations such as a phase shift, as well as changes in trough levels. Elevations of trough concentrations are relatively more pronounced than elevations of mean values over 24 h and have been linked to severity of depression, hence being most consistently observed in melancholic or psychotic major depression (Wong et al. 2000; Gold et al. 2005; Keller et al. 2006).

In this regard, sleep-endocrine research, with a focus on the night period, offers the advantage of fewer confounds (e.g. no food intake) and the added benefit of combining the sleep-EEG with hormone measurements. Since HPA axis activity plays a critical role for the sleep-EEG (Steiger 2002), a combined analysis of ACTH and cortisol secretion on one side and the sleep-EEG on the other, can provide a comprehensive picture of pathophysiological changes associated with major depression. Such analyses, however, require a sophisticated laboratory set-up as well as subjects willing to sleep in a sleep laboratory. Not surprisingly, the number of studies using such an advanced approach, and combining it with a thorough clinical description of patients, is small. The following sections focus on a review of these sophisticated studies, addressing both the methodology as well as the biological and clinical findings.

Alterations of the HPA axis in major depression

An increase in mean 24 h cortisol levels in blood was first described in about half of the patients with endogenous depression, and particularly in elderly patients (Asnis et al. 1981a,b). We also found elevated mean nocturnal cortisol levels particularly in elderly female patients with depression (Antonijevic et al. 2003). This effect of age may have been confounded by the higher chance of elderly patients suffering from melancholic depression (Maes 2002). Indeed, a finer analysis of the cortisol secretion pattern shows that an increase in the trough cortisol concentration is found in particular in melancholic depression, irrespective of age (Wong et al. 2000; Gold et al. 2005; Carroll et al. 2007). A study reporting elevated morning values of ACTH and cortisol specifically in melancholic but not non-melancholic depressed patients is supportive, but less convincing due to the reliance on a single time point measurement in the morning (Kaestner et al. 2005). Our unpublished observation also supports these data in that nocturnal cortisol levels were elevated in non-atypical depression, but were not changed compared to gender- and age-matched controls in patients with atypical depression. Previously, atypical in contrast to melancholic depression has been shown not to be associated with hypercortisolemia (Gold et al. 1995). Also, low CRH concentrations in the CSF have been reported, accompanied by low plasma ACTH levels, suggesting a hypoactive HPA axis (Geracioti et al. 1997). It remains to be explored whether hyperphagia in atypical depression, that is often associated with intake of food rich in fat and carbohydrates, reflects a compensatory mechanism aiming to activate the HPA axis and stimulate serotonin synthesis (Markus et al. 2000; Dallman et al. 2005; Table I).

Table I.  Key studies examining HPA axis activity in the context of further clinical variables to distinguish subgroups within major depression.

Patient cohort/aim of study	Particulars of study	Conclusion/assessment of methodology	References
Eight melancholic patients (6F/2M) vs. 14 controls (6F/6–8M). Evaluation of HPA axis and noradrenergic system activity in melancholia and effect of treatment.	Sophisticated analysis of plasma cortisol and CRH CSF. Due to imperfect gender match between effects analyzed for females separately	Data show elevated plasma noradrenaline and cortisol levels in acute melancholic depression and normalization by treatment. Small patient cohort and focus on female gender. Need for confirmation of key data in another study	Wong et al. (2000) and Gold et al. (2005)
About 12 carefully selected and well-characterized patients. Assessment of cortisol and ACTH secretion pattern in frequent samples over 24 h	Sophisticated cortisol and ACTH secretion pattern analysis. Differentiation of patients according to cortisol secretion into “high” and “low” cortisol patients	Melancholic/psychotic depression associated with elevated plasma cortisol and ACTH, while pulse generation was normal. Small and highly selected patient cohort. Need for confirmation of key data in another study	Carroll et al. (2007)
About 235 acutely depressed inpatients assessed with Dex–CRH test. Assessment of variables affecting the response in the Dex–CRH test	Large sample size allows testing for clinically relevant variables such as number of previous episodes, features of depression and caffeine and nicotine consumption	Vegetative features and HAMD score were positively associated with 15.00 h cortisol levels after dexamethasone. Representative group of inpatients. Confirmation of key findings related to subgroups of patients would be useful. Exploration of findings in outpatients would be relevant	Kunzel et al. (2003)
About 334 patients with acute depression tested with DST and followed up for up to 18 years	DST non-suppression at index episode predicted death by suicide later on in in-patients and those with suicidality	The long-term predictive value of DST non-suppression for suicide risk is substantially enhanced when considering in-patients and those with suicidality. Small number of suicide patients. Need for confirmation of key data in another study	Coryell et al. (2006)
Thirty-seven untreated patients with acute major depression and matched controls. Assessment of morning ACTH and cortisol test in melancholic and non-melancholic patients	Only patients with melancholia had elevated plasma cortisol and ACTH levels compared to controls. Variations in hormone levels were substantial	Data support high HPA axis activity in melancholia, but single time point measurements are not reliable tools to assess HPA activity. Need for confirmation of key data in another study	Kaestner et al. (2005)

Abbreviations: ACTH, adrenocorticotrophic hormone; CRH, corticotropin releasing hormone; CSF, cerebrospinal fluid; Dex, dexamethasone; DST, dexamethasone suppression test; F, female; HAMD, Hamilton rating scale for depression; HPA, hypothalamus–pituitary–adrenal axis; M, male.

A phase—advance of the cortisol and ACTH nadir has been reported in depression by some but not all authors (Linkowski et al. 1987; Rubin et al. 1987b; Antonijevic et al. 2000b; Koenigsberg et al. 2004). The studies reporting a change in cortisol circadian rhythmicity in depression are few, include small sample sizes, and conclusions are based on a non-representative patient group (Linkowski et al. 1987) or on only small phase shifts without pronounced increase in plasma cortisol (Koenigsberg et al. 2004). Therefore, these data do not justify a conclusion about cortisol circadian rhythm changes in a specific subtype of major depression.

Taken together, the data on cortisol and ACTH secretion in depression suggest that compared to age-matched controls melancholic/endogenous major depression is associated with elevated cortisol concentrations, particularly at the time of the cortisol nadir. ACTH levels are also elevated, though less consistently and less significantly. Less data are available for atypical depression, but those published suggest no change or a slight decrease in trough cortisol levels. Since melancholic depression has been reported to have a higher prevalence in older subjects and those with recurrent episodes, the relative impact of these clinical features in addition to the specific biology of melancholia cannot be adequately assessed with the available data.

DST non-suppression occurs in 40–55% of patients with endogenous/melancholic or psychotic depression, with the latter showing the highest rate of non-suppression (up to 64%), while fewer patients with neurotic depression show non-suppression (13–30%) (Rush and Weissenburger 1994; Nelson and Davis 1997; Coryell 2007). Rather, an exaggerated negative feedback response to dexamethasone has been noted (Levitan et al. 2002).

However, the lack of specificity of DST non-suppression and the possible roles of age, anxiety and gender (Greden et al. 1986; Rubin et al. 1987a) do not allow use of this test alone to identify a clinically consistent subtype of major depression. Rather, recent data indicate that DST non-suppression may delineate a relevant biological subgroup only among in-patients with major depression (Coryell et al. 2006; Coryell and Schlesser 2007). Also, it has been suggested that one of the factors determining DST non-suppression is the plasma concentration of dexamethasone (Hermus et al. 1987; Maes et al. 1994), which may also reflect altered metabolism and/or brain penetration in different subtypes of depression.

The Dex–CRH test has been considered a refinement of the DST because an abnormal ACTH and cortisol secretory response has been described in up to 90% of patients (Heuser et al. 1994). This higher sensitivity comes at a cost, namely a lower specificity. Interestingly, a recent study shows, not unexpectedly, that an exaggerated ACTH and cortisol response in the Dex–CRH test is associated with vegetative features (defined as middle and terminal insomnia, weight loss and diurnal variations (Overall and Rhoades 1982)), in elderly (female) patients and recurrent depression (Kunzel et al. 2003). These data support the earlier notion that hypercortisolemia and impaired negative feedback on the HPA axis characterizes a group of depressed patients that best fits the category of melancholic depression. However, a further distinction of this group seems needed, to include recurrence of disease, older age and possibly female gender.

Alterations of the sleep-EEG in major depression

Similar to HPA axis abnormalities, sleep-EEG changes in depression were described many years ago (Kupfer 1971, 1976; Foster et al. 1976; Table II).

Table II.  Key studies examining sleep-EEG changes in the context of further clinical variables to distinguish subgroups within major depression.

Patient cohort/aim of study	Particulars of study	Conclusion/assessment of methodology	References
Thirty-six patients (26F/10M) with depression, 26 of those with familial pure depression, the rest with sporadic depression. Comparison of sleep-EEG at baseline and after 50 mg of amitriptyline	Sophisticated biological characterization of patients (baseline and challenge sleep-EEG) and good clinical characterization. Restriction to conventional sleep-EEG analysis	Baseline differences between groups were small. Amitriptyline induced greater suppression of REM sleep parameters in patients with familial depression. Clinical characterization of patients not commonly used currently. Small number of patients. Need for confirmation of key data in another study	Kupfer et al. (1982)
About 186 patients with major depression, with or without psychotic features. Exploration of prognostic significance of short REM latency. Replication study in 27 psychotic, 79 non-psychotic depressed patients	Sophisticated biological characterization of patients (baseline and challenge sleep-EEG) and good clinical characterization. Restriction to conventional sleep-EEG analysis	Patients with psychotic depression show shortest REM latency and increased wakefulness. These sleep-EEG alterations are more evident in patients with longer illness duration. Similar changes are seen in elderly patients, irrespective of psychotic features. Moderate number of patients with psychotic features examined. Confirmation of key data needed to conclude that psychotic depression is not a distinct biological subgroup	Kupfer et al. (1986) and Thase et al. (1986)
Eight untreated adolescent depressed females and 8 matched controls	Sophisticated biological characterization of patients and comparison with matched controls. Conventional and spectral sleep-EEG analysis	The depressed group showed significantly lower delta power in the first non-REM sleep period. Standard sleep architecture did not differentiate between groups. Small number of patients. Need for confirmation of key data in another study. Need to link delta sleep change to chronicity of depression. Need for exploration of these parameters in male depressed adolescents	Armitage et al. (2001)
Ten patients with atypical depression, with/ without anxiety, in particular panic attacks. Exploration of REM induction by cholinergic stimulation	Sophisticated characterization of patients (atypical depression with and without panic attacks). Clinical relevance supported by treatment studies from same authors. Restriction to conventional sleep-EEG analysis	Data suggest that only patients with atypical features and panic attacks show a biology that is distinct from endogenous depression. Very small patient cohort, particularly when further divided according to anxiety symptoms. Need for confirmation of data in another study	Wager et al. (1990)

Abbreviations: EEG, electroencephalogram; F, female; M, male; REM, rapid eye movement.

REM sleep alterations, the initial focus, were most marked in patients with a positive family history and those with recurrent major depression (Kupfer et al. 1982; Thase et al. 1995a). The disinhibition of REM sleep in depression has been interpreted to reflect a cholinergic hypersensitivity, which can be probed using muscarinic agonists (Krieg and Berger 1987; Poland et al. 1997).

Later, alterations in sleep continuity and non-REM sleep parameters were recognized and have been used to distinguish familial from sporadic forms of major depression (Kupfer et al. 1982). More recent studies have extended the focus to spectral sleep-EEG analysis, showing an association between reduced low frequency delta activity and the risk for recurrence (Kupfer 1995; Buysse et al. 1997).

Interestingly, the finding that among outpatients with depression women did not show reduced SWS and delta activity did not initially receive much attention (Reynolds et al. 1990). We described that both typical REM sleep and SWS/delta activity alterations in women with depression are predominantly seen in elderly, but not pre-menopausal women (Antonijevic et al. 2003). In contrast, our and other data suggest that unlike in women, young and middle-aged men with major depression show typical REM and SWS/delta sleep abnormalities (Armitage et al. 2000; Antonijevic et al. 2000a).

In addition, a shift of delta activity from the first into the second non-REM sleep period has been noted in major depression and has been proposed to signify a risk for recurrence if still present at remission (Kupfer 1995; Buysse et al. 1997). However, the biological basis of an altered SWS and delta activity pattern in depression remains unclear. One possible explanation, based on data with an experimental antidepressant, could be that the abnormal shift in delta activity reflects either reduced serotonin neurotransmission or up-regulation of dorsal raphe 5-HT1A auto-receptors or both (Murck et al. 2001). This hypothesis is supported by a study in depression showing a normalization of the delta sleep pattern by treatment with a selective serotonin-reuptake inhibitor (Jindal et al. 2003).

Interestingly, a preliminary and hence small study has reported primarily reduced delta activity in the first non-REM sleep cycle in female adolescents with depression (Armitage et al. 2001). These data open up the hypothesis that in early onset depression, subtle sleep-EEG changes are present, that are only detected using sophisticated spectral, but not conventional sleep-EEG analysis.

In our hands, sleep-EEG alterations in adult women and men, analyzed according to the presence or absence of atypical or features, demonstrated that women with atypical features have no sleep-EEG alterations considered typical for major depression except for an altered delta sleep pattern. In contrast, men with atypical depression showed an altered delta sleep pattern as well as sleep-EEG abnormalities which were merely less pronounced compared to patients with non-atypical features (Antonijevic, unpublished observation). These data support a critical role of gender with regard to the biology of atypical features. On the other hand, these data also question the current clinical criteria for atypical depression, as these do not seem to delineate a biologically homogenous group of patients, particularly when combining men and women (Posternak and Zimmerman 2001, 2002; Parker et al. 2002, 2005).

Since patients without typical sleep-EEG alterations respond more favorably to interpersonal psychotherapy than depressed patients with typical sleep-EEG alterations (Thase et al. 1997), absence or presence of such sleep-EEG measures could have major treatment implications. As no data have been reported in patients with depression that show only an abnormal delta sleep ratio, it would be very worthwhile to explore in future prospective studies whether in such patients normalization of the delta sleep pattern, either by psychotherapy or pharmacological treatment, is associated with long-term recovery. This would be particularly relevant for a very young patient population with few biological alterations, but a high vulnerability to develop chronic depression if not adequately treated.

Alterations of the HPA axis and concomitant changes of the sleep-EEG in major depression

In a complex albeit small study Hatzinger et al. combined the response in the Dex–CRH test with sleep-EEG parameters to determine the short-term response to treatment as well as the long-term course of disease (Hatzinger et al. 2002, 2004), lower ACTH and cortisol responses in the Dex–CRH test were associated with early improvement on treatment. The cortisol response in the Dex–CRH test was most consistently and inversely correlated with sleep continuity variables. At week 6 (on treatment), the cortisol response was also negatively correlated with the duration of SWS. In contrast, REM sleep parameters, with the exception of REM density, showed no association with the cortisol response. These data are in agreement with other authors, suggesting that REM sleep parameters represent trait rather than state markers of recurrent depression. Interestingly, the inverse correlation between cortisol and SWS after 6 weeks of treatment could mark out a group of patients with a risk for recurrence, if SWS and cortisol levels are not normalized (Table III).

Table III.  Key studies examining HPA axis activity and sleep-EEG changes in the context of further clinical variables to distinguish subgroups within major depression.

Patient cohort/aim of study	Particulars of study	Conclusion/assessment of methodology	References
About 44 psychotic and 44 non- psychotic untreated depressed patients (30%M/70%F in both groups). Assessment of sleep- EEG and DST. Greater severity of depression in psychotic patients	Sophisticated biological and clini cal characterization of patients. Restriction to conventional sleep-EEG analysis	Psychotic depression was pri marily associated with short REM latency, while other sleep-EEG changes and hypercortisolism were associated with severity of depression and age. Moder ate number of patients with psychotic features. The data extend and confirm previous findings that psychotic depression is characterized by a short REM latency, but is otherwise not a distinct biological subgroup	Stefos et al. (1998)
About 10–15 depressed patients (both M/F) assessed by repeated sleep-EEG analysis and Dex– CRH test before and on treat ment. Correlation between bio logical measures	Sophisticated biological characterization of patients with follow-up data on treatment. Restriction to conventional sleep-EEG analysis	Measures of HPA overdrive were most strongly associ ated with reduced sleep continuity. SWS amount was moderately associated with HPA activity. No detailed clinical charac terization of patients. Small number of patients. Need for confirmation of key data in another study	Hatzinger et al. (2002, 2004)
About 16 F inpatients patients, 19 F controls. Exploration of age and menstrual status on sleep- endocrine changes in major depression	Sophisticated biological characterization of patients and comparison with matched controls. Conventional and spectral sleep-EEG analysis	Enhanced nocturnal HPA activity and typical sleep alterations seen in postmenopausal depressed patients. Young female patients show only abnormal delta sleep ratio. No detailed clinical characterization of patients. Small number of patients. Need for confir mation of key data in another study	Antonijevic et al. (2003)
Eight depressed patients (5F/3M) and matched 8 controls. Exploration of sleep-EEG and single time point nocturnal cortisol measurement after muscarinic agonist (RS 86)	Sophisticated biological characterization of patients and comparison with matched controls. Restriction to con ventional sleep parameters. Small sample size may have precluded differ ence in response to cholinergic stimu lation between patients and controls	Small number of patients. No detailed clinical characteriz ation of patients. Data do not allow identification of differ ent biological and clinical subgroups of major depression.	Krieg and Berger (1987)

Abbreviations: Dex–CRH, dexamethasone–corticotropin releasing hormone; DST, dexamethasone suppression test; EEG, electroencephalogram; F, female; HPA, hypothalamus–pituitary–adrenal axis; M, male; REM, rapid eye movement; SWS, slow wave sleep.

These data are also backed up by our previous study showing that nocturnal cortisol concentrations are inversely correlated with the amount of non-REM sleep (Antonijevic et al. 2000b). These data lend further support to the hypothesis that melancholic depression is characterized by enhanced activity of systems that cause arousal (elevated noradrenergic tone and hypothalamic CRH secretion), while atypical depression is associated with reduced activity of these systems (Gold and Chrousos 2002).

Interestingly, it has been suggested that hypercortisolemia, in the context of pro-inflammatory cytokine secretion, is associated with low sleep efficiency (Vgontzas et al. 2006). An interaction between inflammatory cytokines and HPA activity in major depression has been suggested for several years (Anisman et al. 1999; Wichers and Maes 2002; Leonard 2005). However, this complex topic clearly extends beyond this review. Therefore, I should only point out that inflammatory cytokines could prove relevant for the subtype-specific differences in sleep-endocrine changes in major depression.

Inflammatory mediators could also be relevant to further clarify the impact of gender on stress-endocrine alterations in depression, which currently is far from understood: while onset of depression in young women seems to be associated with a chronic course, atypical features, low cortisol concentrations and no sleep-EEG abnormalities, depression in elderly women, particularly after the menopause, has been associated with high cortisol concentrations and multiple typical sleep-EEG alterations (Antonijevic et al. 2003). In men, such a clear impact of age has not been observed. Our unpublished observations rather show that men with atypical features have more sleep-EEG changes than women with atypical depression, but less pronounced sleep-EEG changes compared to men with non-atypical depression. As nocturnal cortisol levels were lower in both men and women with atypical features compared to patients with non-atypical features, our data suggest that sleep-EEG alterations and HPA axis changes are not necessarily tightly correlated, at least in depression with atypical features.

Conclusions and outlook

In the past decades, several studies have shown that biological differences among patients with major depression can be identified. Sleep-EEG and stress-hormone assessments are the most widely studied tools to differentiate subtypes of major depression and expose the underlying biology. Despite the many studies a firm demarcation of biologically and clinically identifiable patient groups is still lacking. Factors contributing to the difficulties in reaching conclusions about biologically defined subtypes include (1) the change in clinically-defined diagnostic criteria in 1980 and the continued debate about criteria for melancholic and atypical depression, (2) the often small sample sizes used to perform complex biological measures, such as sleep-EEG and repeated HPA hormone measurements, and (3) the lack of unanimously agreed fine details of the biological read-outs. As an example of the third point, dexamethasone has been used at different doses, without consistent measurement of dexamethasome blood levels, and measurements of cortisol at different time points.

Without agreement on and consistent use of defined clinical criteria and biological assessments independent replication of data is hampered. Replication and confirmation of findings, however, is a prerequisite to advance our understanding of the biology of the emerging subtypes of major depression.

Currently available data highlight the relevance of age, gender, recurrence vs. chronicity of symptoms and in-patient vs. out-patient status for the interpretation of biological alterations. Also, a familial aggregation of severe and/or melancholic depression seems to affect the risk in the offspring for severe and melancholic depression, irrespective of gender (Schreier et al. 2006).

Furthermore, recent studies point to the importance of less tangible moderators of the biology of major depression, such as life events. Though inclusion of these data goes beyond the scope of this review, it seems critical to at least mention that the frequency of recent life events has been positively associated with atypical, but not with melancholic depression (Fountoulakis et al. 2006; Coryell 2007).

Such data bring up the interesting but so far unresolved matter to what extent some forms of depression are a reaction to (emotional) stress. Animal as well as human research converges on the point that early life experiences shape HPA responsiveness to stressful situations in adult life (Heim et al. 2004; Pruessner et al. 2004; Meaney and Szyf 2005; Weaver et al. 2006). In combination with a specific genotype, such stressful events can markedly increase the risk to develop depression, particularly in women (Caspi et al. 2003; Kendler et al. 2005). It remains to be explored whether these patients can be biologically characterized using sleep-EEG analysis and HPA axis function tests.

Finally, it is pertinent to stress that depressive disorders with atypical and melancholic features represent the opposite ends of a spectrum of disorders. Thus, the spectrum of bipolar disorders, which in the case of bipolar II seems to also overlap with atypical depression (Posternak and Zimmerman 2002), and depressive disorders in the context of anxiety disorders, need to be comprehensively explored in terms of sleep-EEG and stress-hormone alterations. This is particularly relevant as some data point to a sustained HPA overdrive in patients with bipolar disorder even during remission (Watson et al. 2006).

Moreover, as mentioned before, the clinical criteria for melancholic and atypical depression are still being debated. Thus, Parker has been an advocate for inclusion of psychomotor retardation to delineate a biologically more homogenous group of patients presenting with melancholic or psychotic features (Parker et al. 2003). It would be most valuable to explore prospectively whether the specificity of DST non-suppression can be augmented by testing patients with currently defined melancholic features with and without concomitant psychomotor retardation.

Unlike melancholia, which is an identifiable clinical syndrome despite the debate on the definite clinical criteria, the clinical description of atypical depression remains more elusive. Thus, the mandatory requirement for mood reagibility has been questioned, as it seems to be associated with less severe depression rather than with a specific subtype (Parker et al. 2002). On the other hand, the chronic course and the early onset seem to define a biologically more homogenous group of patients with atypical features (Stewart et al. 2005). The clinical relevance of such distinction is underscored by the differential treatment response (Quitkin et al. 1989, 1993; Stewart et al. 1998) and the stability of clinical features, at least for two consecutive depressive episodes (Nierenberg et al. 1996).

The inclusion of age of onset and course of disease may also help identify consistent biological markers in atypical depression, which so far is mostly defined by lack of alterations typically associated with major depression. Also, the use of more sophisticated, but not necessarily more complex, assessments such as the delta sleep ratio or delta sleep during the first sleep cycle may prove to be valuable tools to understand the biology of this poorly understood clinical syndrome.

Also, a re-emphasis of reversed vegetative features, namely hypersomnia and hyperphagia, in the context of onset and course of disease, may help to better define a biologically homogenous group of patients than the current broader criteria of atypical depression.

The better endophenotyping of patients with depressive disorders should also help to recognize patient groups that are currently often inadequately diagnosed and treated, such as patients with apparent stress-related depressive disorders. A recent study serves as a good example: patients diagnosed with different depressive disorders in response to job stress were shown to share an attenuated cortisol response in the dexamethasone–CRH test and some clinical features (Rydmark et al. 2006). Independent replications of such associations between clinically relevant contextual and biological features are needed to advance our understanding of depression with its multifaceted pathophysiology. Ultimately, this approach will not only achieve higher remission rates, but will also facilitate the discovery of novel drug targets for distinct subtypes of major depression.