Differentiating melancholic and non-melancholic depression via biological markers: A review

Figures & data

Table 1. Search strategy for MEDLINE, PsycINFO, and EMBASE.

Specific subject headings PsycINFO: exp endogenous depression/ Embase: exp melancholia/ OR exp endogenous depression/ (depress* OR melanchol*).ti. Broad subject headings PsycINFO: exp major depression/ OR exp affective psychosis/ MEDLINE: exp depressive disorder/ OR exp depressive disorder, major/ OR affective disorders, psychotic/ Embase: exp major depression/ OR exp depressive psychosis/ 2 OR 3 (Melanchol* OR endogenous?? depress* OR (endogenous?? AND (non-endogenous?? depress* OR nonendogenous?? depress*)) OR psychotic depress* OR (psychotic AND (non-psychotic depress* OR nonpsychotic depress*)) OR typical depress* OR (typical AND (non-typical depress* OR nontypical depress*)) OR nonatypical depress* OR non-atypical depress* OR ((nonatypical OR non-atypical) AND atypical depress*) OR endogenomorphic depress* OR vital depress* OR type A depress* OR autonomous depress*).ti,ab. (Type? of depress* OR type? in depress* OR depression type? OR depressive type? OR depressed type? OR (depress* AND (subtyp* OR sub-typ* OR subgrou* OR sub-grou*))).ti. 4 AND (5 OR 6) (Depress* AND (endogenous features OR psychotic features OR typical features OR nonatypical features OR non-atypical features OR endogenomorphic features OR vital features OR type A features OR autonomous features)).ti. (Endogenous?? unipolar depress* OR psychotic unipolar depress* OR typical unipolar depress* OR nonatypical unipolar depress* OR non-atypical unipolar depress* OR endogenomorphic unipolar depress* OR vital unipolar depress* OR type A unipolar depress* OR autonomous unipolar depress*).ti,ab. (Endogenous?? major depress* OR psychotic major depress* OR typical major depress* OR nonatypical major depress* OR non-atypical major depress* OR endogenomorphic major depress* OR vital major depress* OR type A major depress* OR autonomous major depress*).ti,ab. (Endogenous?? unipolar major depress* OR psychotic unipolar major depress* OR typical unipolar major depress* OR nonatypical unipolar major depress* OR non-atypical unipolar major depress* OR endogenomorphic unipolar major depress* OR vital unipolar major depress* OR type A unipolar major depress* OR autonomous unipolar major depress*).ti,ab. 1 OR 7 OR 8 OR 9 OR 10 OR 11 12 AND bio*.mp. limit 13 to yr=‘2005-current’

Figure 1. PRISMA flow diagram outlining the search strategy and study selection process.

Table 2. Definitional criteria for melancholia.

Criteria	Summary
Diagnostic and Statistical Manuel of Mental Disorders (DSM)	Three versions covered in studies: DSM-III, DSM-III-R, DSM-IV, and DSM-5. Most revolve around the following symptoms: Loss of pleasure in all or almost all activities. Lack of reactivity to usual pleasurable stimuli (does not feel much better, even temporarily, when something good happens). Distinct quality to depressed mood (i.e. depressed mood is perceived as distinctly different from the kind of feeling patient would have or has had following the death of a loved one). The depression is regularly worse in the morning. Early-morning awakening (at least 2 hours before usual time of awakening). Marked psychomotor retardation or agitation. Significant anorexia or weight loss. Excessive or inappropriate guilt. DSM-III melancholia requires both (1) and (2), and at least three of (3)–(8). DSM-IV and DSM-5 are similar, but only require one of (1) or (2). DSM-5 specifies that (3) can be characterised by profound despondency, despair, and/or moroseness or by so-called empty mood. DSM-III-R requires at least 5 symptoms from a list consisting of (1)–(2), (4)–(7), and three other symptoms: No significant personality disturbance before first major depressive episode One or more previous major depressive episodes followed by complete or nearly complete recovery. Previous good response to specific and adequate somatic antidepressant treatment
Newcastle Index of Endogenous Depression (Carney et al. 1965)	Ten-item index with differing weights (given in brackets). Endogenous depression suggested to be indicated by a score of 6 or more. Items Adequate personality (i.e. no history of neurosis or social maladjustment, +1), no adequate psychogenesis (+2), distinct quality (i.e. reaction to adversity that is not ‘normal’ sadness or gloom, +1), weight loss (+2), previous depressive episode (+1), depressive psychomotor activity (+2), anxiety (–1), nihilistic delusions (+2), blame others (–1), guilt (+1)
Research Diagnostic Criteria (Spitzer et al. 1978)	Four symptoms of the below for probable endogenous depression, and six for definite. At least one symptom must come from Group A. Group A Distinct quality to depressed mood, i.e. depressed mood is perceived as distinctly different from the kind of feeling patient would have or has had following the death of a loved one. Lack of reactivity to environmental changes (once depressed, does not feel better, even temporarily, when something good happens). Mood is regularly worse in the morning. Pervasive loss of interest or pleasure (some loss in all areas). Group B Feelings of self-reproach or excessive or inappropriate guilt. Early-morning awakening or middle insomnia. Psychomotor retardation or agitation (more than mere subjective feeling of being slowed down or restless). Poor appetite. Weight loss (2 kg/wk over several weeks or 10 kg/year when not dieting). Loss of interest or pleasure (may or may not be pervasive) in usual activities or decreased sexual drive.
The CORE Measure (Parker et al. 1990, 1994)	Composite measure of 2 groups of 18 signs (predominantly psychomotor features) and symptoms. Signs Non-interactiveness, facial immobility, postural slumping, non-reactivity, facial apprehension, delay in responding verbally, shortened verbal responses, inattentiveness, facial agitation, body immobility, motor agitation, poverty of associations, slowed speed of movement, verbal stereotypy, delay in motor activity, impaired spontaneity of talk, slowing of speech rate, stereotyped movements. Symptoms Appetite loss, weight loss, slowed thinking, indecisive, unpleasant thoughts, slowed physically, suicidal thoughts, loss of interest in pleasurable activities, anticipatory anhedonia, consummatory anhedonia, incapacity to be cheered by a pleasant event, incapacity to be cheered by social support, mood worse in morning, energy worse in morning, terminal insomnia, non-variable mood, constipation, pathological guilt.
The Michigan Index (Feinberg and Carroll 1982)	An index of 7 items of differing weights and scoring ranges (given in brackets). A score of >26 or above indicates endogenous depression, 19–26 is unclear, and <19 is non-endogenous. Symptoms Decreased appetite (weight 9; range 0–2), guilt (6; 0–4), psychomotor agitation (4; 0–4), affective delusions (3; 0–8), loss of interest in/capacity for work or activities (3; 0–4), psychomotor retardation (2; 0–4), loss of pleasure (2; 0–2).

Table 3. Summaries of studies exploring endocrinological markers of melancholia.

Study	Sample size	Age (mean (SD))	Sex (% female)	MEL diagnostic criteria	Depression severity (mean (SD))	Measurement method	Results	Notes
Dexamethasone Suppression Test (DST)
Diagnostic test metrics
Brown et al. 1988	93 (M = 73)	60 (14)	66	DSM-III	HDRS-24 M = 27.8 (6) MD = 37 (9.5) MA = 31.8 (6.2) MDA =37.6 (8.9) N = 22.8 (5)	Plasma cortisol non-suppression (for M)	TPR = 0.71 TNR = 0.95 PPV = 0.98 NPV = 0.48 Acc = 0.78	Optimal post-dexamethasone plasma cortisol cut-off identified by Carroll et al. (1981) also replicated.
Carroll et al. 1981	368 (M = 215)	48 (16)	M = 64 N = 61	27% Clinician 73% Clinician consensus based off SADS and RDC	CRSD M = 25.7 (9.7) N = 23.7 (9.5) HDRS M = 19.9 (7.9) N = 16.2 (7.5)	Plasma cortisol non-suppression (for M)	TPR = 0.67 TNR = 0.96	Seminal study; established optimal dexamethasone dose (1 mg) post-dexamethasone plasma cortisol (5 μg/dL). Non-suppression was unrelated to potentially confounding variables such as age, sex, medication history or symptom severity. 4% Of variance in post-dexamethasone plasma cortisol was accounted for by HDRS severity.
Davidson et al. 1984	49 (M = 16)	M = 50 (7) N = 44 (6)	Not reported	NEDDI RDC DSM-III Michigan	HDRS-17 Suppressors = 31 (5) Non-suppressors = 26 (5)	Plasma cortisol non-suppression (for M)	DSM-III TPR = 0.45 TNR = 0.80 PPV = 0.66 RDC TPR = 0.40 TNR = 1.00 PPV = 1.00 Newcastle TPR = 0.48 TNR = 0.89 PPV = 0.87 Michigan TPR = 0.39 TNR = 0.89 PPV = 0.94	Contrasted four different melancholic definitions. DSM-III assignments performed the worst while RDC, NEDDI and Michigan definitions performed quite similarly.
Georgotas et al. 1986	66 (M = 46/36) (RDC/NEDDI)	67.41 (7.07) RDC M = 67.65 (7.19) N = 66.85 (6.89) NEDDI M = 67.94 (7.56) N = 66.77 (6.46)	65.15 RDC M = 56.52 N = 85 NEDDI M = 58.33 N = 73.33	RDC NEDDI	HDRS-21 RDC M = 26.67 (7.73) N = 25.00 (5.81) NEDDI M = 27.69 (8.22) N = 73.33 (5.26)	Plasma cortisol non-suppression (for M)	RDC TPR = 0.46 TNR = 0.80 NEDDI TPR = 0.58 TNR = 0.87	Also provided sensitivity and specificity for various suppression cut-offs from 3-12μg/dL.
Holden 1983	41 (M = 22)	55.5	70.7	NEDDI	Not reported	Plasma cortisol non-suppression (for M)	TPR = 0.82 TNR = 0.89 PPV = 0.94	‘Endogenous’ versus ‘non-endogenous’ depression.
Miller and Nelson 1987	84 (M = 45)	49 (19)	71.6	DSM-III RDC	Not reported	Serum cortisol non-suppression (for M)	See notes	DSM-III-defined melancholia and/or endogenous depression were not associated with DST non-suppression. Vegetative symptoms like insomnia, agitation, weight loss, and lack of libido were linked to a greater likelihood of non-suppression. Reported age and sex were for larger sample consisting of a further 11 patients with alternative diagnoses.
Rush et al. 1982	70 (M = 32)	36.5 (11.3)	Roughly two thirds	RDC (SADS-L)	HDRS-17 M = 26.7 (5.1) N = 21.4 (5.8) BDI M = 30.6 (8.3) N = 27.5 (9.5)	Serum cortisol non-suppression (for M)	TPR = 0.41 TNR = 0.95
Non-suppression rates
Amsterdam et al. 1989 (Study 1)	72 (M = 51)	M = 38.33 (12.71) N = 34.43 (9.74)	M = 66.67 N = 71.43	DSM-III	Not reported	Serum cortisol non-suppression rates	M > N	M = 43.1%, N = 4.8%. Significant from self-calculated χ² test.
Amsterdam et al. 1989 (Study 5)	32 (M = 22)	39.13 (13.04)	71.88	DSM-III	Not reported	Serum cortisol non-suppression rates	M > N	M = 54.5%, N = 10%. Significant from self-calculated Fisher’s exact test.
Brown and Shuey 1980	24 (M = 6)	Not reported	Not reported	RDC	Not reported	Serum cortisol non-suppression rates	M > N	M = 66.67%, N = 50%. Not significant from self-calculated Fisher’s exact test. Most of the study only reported differences between suppressors and non-suppressors irrespective of diagnosis. But reported higher rates of non-suppression in MEL compared to NMD patients.
Brown et al. 1988	93 (M = 73)	60 (14)	66	DSM-III	HDRS-24 M = 27.8 (6) MD = 37 (9.5) MA = 31.8 (6.2) MDA =37.6 (8.9) N = 22.8 (5)	Plasma cortisol non-suppression rates	M > N	M = 86.3%, N = 23.5%. Significant from self-calculated χ² test. Non-suppression was more likely in agitated or delusional MEL patients, but this group was small.
Evans and Nemeroff 1987	85 (M = 23)	37.7	55.42	DSM-III	Not reported	Serum cortisol non-suppression rates	M > N	M = 78%, N = 48%. Also reported 95% for 19 with psychotic depression (unclear whether mutually exclusive from melancholia) and various suppression cut-offs from 4-15μg/dL. Reported age and sex were for larger sample consisting of a further 45 patients with alternative diagnoses.
Georgotas et al. 1986	66 (M = 46/36) (RDC/NEDDI)	67.41 (7.07) RDC M = 67.65 (7.19) N = 66.85 (6.89) NEDDI M = 67.94 (7.56) N = 66.77 (6.46)	65.15 RDC M = 56.52 N = 85 NEDDI M = 58.33 N = 73.33	RDC NEDDI	HDRS-21 RDC M = 26.67 (7.73) N = 25.00 (5.81) NEDDI M = 27.69 (8.22) N = 73.33 (5.26)	Plasma cortisol non-suppression rates	RDC M > N NEDDI M > N	RDC M = 46%, N = 20%, NEDDI M = 58%, N = 13%. Results obtained from reported χ² tests.
Peselow and Fieve 1992	85 (M = 42)	M = 47.43 (12.5) N = 41.30 (12.8)	M = 30.95 N = 44.19	Both RDC (SADS-C) and DSM-III	HDRS M = 28.83 (5.7) N = 26.33 (5.0) BDI M = 28.33 (4.3) N = 27.56 (3.3) CGI M = 4.79 (0.8) N = 4.33 (0.6) Raskin M = 11.31 (1.6) N = 10.77 (1.2) MADRS M = 34.21 (5.9) N = 31.26 (5.3)	Plasma cortisol non-suppression rates	M > N	Individuals who met two definitions of melancholic/endogenous depression (i.e. RDC and DSM-III) showed 64.3% non-suppression, compared to 23.3% who met neither definition. Significant from self-calculated χ² test.
Rush et al. 1996	422 RDC M = 184 DSM-III M = 69	RDC M = 37.5 (13.6) N = 36.4 (11.4) DSM-III M = 40.9 (16.0) N = 35.7 (11.0)	RDC M = 60.3 N = 68.5 DSM-III M = 54.4 N = 64.9	RDC DSM-III	HDRS-17 RDC M = 26.3 (6.2) N = 19.8 (5.0) DSM-III M = 28.3 (6.6) N = 21.2 (5.8)	Serum cortisol non-suppression rates, cut-off 4 μg/dL	RDC M > N DSM-III M > N	RDC M = 39.1%, N = 9.7%, DSM M = 40.6%, N = 16.5%. Both significant according to self-calculated χ² tests. Also reports suppression rates at peak cortisol level which has similar results.
Rush et al. 1997	133 (M = 85)	M = 39.2 (13.1) N = 36.2 (10.2)	M = 65.9 N = 60.4	RDC (SADS-L)	HDRS-17 M = 28.3 (8.5) N = 18.7 (4.1)	Serum cortisol non-suppression rates	M > N	M = 45.9%, N = 14.6%. Significant according to self-calculated χ² tests.
Post-dexamethasone cortisol levels
Georgotas et al. 1986	66 RDC M = 46 NEDDI M = 36	67.41 (7.07) RDC M = 67.65 (7.19) N = 66.85 (6.89) NEDDI M = 67.94 (7.56) N = 66.77 (6.46)	65.15 RDC M = 56.52 N = 85 NEDDI M = 58.33 N = 73.33	RDC NEDDI	HDRS-21 RDC M = 26.67 (7.73) N = 25.00 (5.81) NEDDI M = 27.69 (8.22) N = 73.33 (5.26)	Plasma cortisol, μg/dL	RDC M > N NEDDI M > N	RDC M = 5.80 (5.43) N = 3.83 (3.62) NEDDI M = 7.00 (5.58) N = 3.05 (3.18) Results obtained from ANCOVA controlling for severity of depression.
Hubain et al. 1998	300 (M = 95)	44 (12)	66	NEDDI	HDRS-24 Female = 27 (7) Male = 29 (9)	Plasma cortisol, μg/L	M > N	M = 36.5, N = 22.5 (geometric means; t-test reported significant). No DST results were influenced by gender.
Maes, Maes et al. 1992	80 M = 35	M = 53.1 (13.02) N = 41.4 (12.75)	0	Cluster analysis	HDRS M = 26.9 (5.32) N = 18.7 (4.02)	Plasma cortisol, no units given	M > N	Looks at ‘vital’ (which is 80% M according to DSM-III-R) and ‘non-vital’ depression. Identified symptom profiles consistent with melancholic and non-melancholic depression. Significant difference reported with ANOVA.
Maes, Meltzer, et al. 1994	45 (M = 21)	M = 51.7 (9.62) N = 45.3 (13.23)	Not reported	DSM-III (SCID)	HDRS-17 M = 26.2 (2.75) N = 21.8 (2.45)	Plasma levels, μg/dL	M > N	M = 11.7 (7.79), N = 5.8 (5.39). Significant according to self-calculated Welch’s test.
Maes, Scharpé, et al. 1993	17 (M = 10)	53.5 (14.6)	Not reported	DSM-III (SCID)	HDRS-17 M = 21.6 (3.8) N = 27.5 (3.2)	Serum levels, μg/dL	M > N	M = 7.6 (5.9), N = 4.0 (3.7). Non-significant according to self-calculated Welch’s test.
Rush et al. 1996	422 RDC M = 184 DSM-III M = 69	RDC M = 37.5 (13.6) N = 36.4 (11.4) DSM-III M = 40.9 (16.0) N = 35.7 (11.0)	RDC M = 60.3 N = 68.5 DSM-III M = 54.4 N = 64.9	RDC DSM-III	HDRS-17 RDC M = 26.3 (6.2) N = 19.8 (5.0) DSM-III M = 28.3 (6.6) N = 21.2 (5.8)	Serum levels at highest point (H) and at 4 pm after administration at 11 pm (14h), μg/dL	RDC H M > N 14h M > N DSM-III H M > N 14h M > N	See Tables 3 and 5; DSM H N = 2.8 (3.0), 14h N = 2.5 (2.6). RDC diagnostic differences reported significant, DSM confirmed through self-calculated Welch’s tests.
Türkçapar et al. 1999	65 RDC M = 43 DSM-III M = 16 DSM III-R M = 32 DSM-IV M = 37 ICD-10 M = 41	Overall 33.3 (9.9) RDC M = 35.2 (9.8) N = 29.6 (9.3) DSM-III M = 35 (11.1) N = 32.6 (10.8) DSM III-R M = 37 (10.3) N = 28 (7.3) DSM-IV M = 35.3 (9.9) N = 30.7 (9.4) ICD-10 M = 35.6 (10.3) N = 29.4 (7.9)	63.1	RDC DSM-III (SCID) DSM III-R DSM-IV ICD-10	HDRS-17 Overall 24.0 (6.3) RDC M = 26.2 (6.1) N = 20.0 (4.5) DSM-III M = 26.8 (5.2) N = 23.2 (6.4) DSM III-R M = 25 (6.7) N = 23.1 (5.9) DSM-IV M = 26.2 (6.8) N = 21.3 (4.4) ICD-10 M = 25.9 (6.6) N = 21.0 (4.4)	Plasma levels, μg/dL	RDC M > N DSM-III M > N DSM III-R M > N DSM-IV M > N ICD-10 M > N	See Tables 2 and 3 in study. T-tests reported. Those with melancholia defined by DSM-III or DSM-III-R criteria did not display significant post-DST cortisol differences (most restrictive criteria) compared to NMD patients, but that those with DSM-IV-defined melancholia did. RDC least restrictive tool.
Natural cortisol levels
Amsterdam et al. 1989 (Study 1)	72 (M = 51)	M = 38.33 (12.71) N = 34.43 (9.74)	M = 66.67 N = 71.43	DSM-III	Not reported	Serum cortisol, μg/dL	M > N	M = 14.4 (4.5), N = 12.6 (7.3). Non-significant from self-calculated Welch’s test.
Amsterdam et al. 1989 (Study 2)	26 (M = 14)	35.7 (9.6)	Not reported	DSM-III	Not reported	Plasma cortisol, μg/mL	M > N	M = 15.7 (5.1), 14.4 (6.3). Non-significant from self-calculated Welch’s test.
Amsterdam et al. 1989 (Study 5)	32 (M = 22)	39.13 (13.04)	71.88	DSM-III	Not reported	Serum cortisol, μg/dL	M > N	M = 14.0 (3.8), N = 11.8 (6.6). Non-significant from self-calculated Welch’s test.
Brouwer et al. 2005	113 M = 32 N1 = 25 N2 = 56	47 (11)	62	DSM-IV (SCID)	HDRS-17 21 (3)	Serum cortisol (nmol/L), urinary cortisol (nmol/24h)	Serum M > N1 N2 > M Urinary M > N1 M > N2	N1 = atypical depression, N2 = unspecified depression. Reported p values showed no differences between MEL and either NMD groups for either serum or urinary cortisol.
Cizza et al. 2012	41 (M = 31)	M = 34.3 (7.0) N = 36.6 (6.8)	100	DSM-IV (SCID)	HDRS M = 8.1 (5.3) N = 10.6 (8.1)	Plasma cortisol (μg/dL)	M > N	All N are atypical. Age and HDRS reported for whole sample (n = 89, 51 MEL). Non-significance reported.
Herane-Vives et al.  2020	71 M = 19	M = 38.2 (13.1) N = 31.8 (8.9)	M = 52.6 N = 73.1	DSM-5 (MINI) NEDDI	HDRS-17 M = 17.6 (6.5) N = 17.2 (4.8)	Hair cortisol concentration (HCC; pg/mg) Serum cortisol levels (nmol/l.h), AUCg, cortisol awakening response (CAR), 30 min delta cortisol secretion after awakening (DELTA)	HCC M > N AUCg M > N CAR N > M DELTA N > M	See Supplementary Table 2 from study. All comparisons non-significant from reported ANOVA (included controls) p-values. AUCg significant omnibus test but post-hoc M v N insignificant. Hair and salivary cortisol used as measures of long- and short-term cortisol levels respectively. In NMD patients only, decreased early morning delta cortisol (the difference between peak and basal cortisol levels within a 30-minute period after wakening) was associated with higher depressive severity. Both MEL and NMD patients who showed lower long-term cortisol levels also tended to have higher depressive severity.
Karlović et al. 2012	55 (M = 32)	M = 48.6 (8.1) N = 50.9 (8.3)	M = 37.5 N = 34.8	DSM-IV-TR (SCID)	HDRS-17 M = 30.1 (10.7) N = 21.5 (7.9)	Serum cortisol levels, nmol/L	M > N Positive association	M = 822.1 (43.8), N = 557 (55.9). N = atypical depression. Significant from self-calculated Welch test. Also positive association with M in regression model controlling for demographics, IL-6, CRP, and TNFα. β = 0.011, p = 0.002. DSM-IV-TR-defined MEL patients had higher cortisol levels than patients with atypical depression. No other significant differences were found.
Lamers et al. 2013	233 (M = 111)	M = 40.2 (12.1) N = 39.6 (12.1)	M = 65.8 N = 79.5	Latent class and latent transition analysis	IDS-30 M = 39.1 (10.0) N = 35.2 (11.5)	Salivary cortisol wakening curves, μg/dL Area under the curve with respect to ground (AUCg) and increase (AUCi), diurnal slope (mean decline per hour)	AUCg M > N AUCi M > N Slope M > N	N = atypical depression. Described as severe subtypes. Reported pairwise comparison p-values adjusted for sex, age, education level, smoking, and awakening time showed MEL higher for slope and AUCg.
Michopoulos et al. 2008	40 M = 20	M = 55.1 (10.1) N = 50.3 (11.1)	100	DSM-IV-TR (SCID)	HDRS-17 M = 20.2 (3.2) N = 19.8 (4.8)	Plasma cortisol (mg/dL) Salivary cortisol (ig/mL), measured in morning (AM), noon, and night (PM)	Plasma M > N Saliva AM M > N Saliva Noon M > N Saliva Cortisol M = N	See Table 2 of study. All cortisol measures insignificant by reported t or Mann–Whitney tests. MEL patients showed a strong positive correlation between noon salivary cortisol levels and performance on the Stockings of Cambridge (SOC) task of executive function, while NMD patients showed a negative correlation between cortisol plasma and SOC performance. NMD patients also showed a negative correlation between plasma cortisol and stages completed on an attentional set shifting task, while showing strong positive correlations between two measures of error rates on the same set shifting task.
Türkçapar et al. 1999	65 RDC M = 43 DSM-III M = 16 DSM III-R M = 32 DSM-IV M = 37 ICD-10 M = 41	Overall 33.3 (9.9) RDC M = 35.2 (9.8) N = 29.6 (9.3) DSM-III M = 35 (11.1) N = 32.6 (10.8) DSM III-R M = 37 (10.3) N = 28 (7.3) DSM-IV M = 35.3 (9.9) N = 30.7 (9.4) ICD-10 M = 35.6 (10.3) N = 29.4 (7.9)	63.1	RDC DSM-III (SCID) DSM III-R DSM-IV ICD-10	HDRS-17 Overall 24.0 (6.3) RDC M = 26.2 (6.1) N = 20.0 (4.5) DSM-III M = 26.8 (5.2) N = 23.2 (6.4) DSM III-R M = 25 (6.7) N = 23.1 (5.9) DSM-IV M = 26.2 (6.8) N = 21.3 (4.4) ICD-10 M = 25.9 (6.6) N = 21.0 (4.4)	Plasma levels, μg/dL	RDC M > N DSM-III M > N DSM III-R M > N DSM-IV M > N ICD-10 M > N	See Tables 2 and 3 in study. T-tests reported. RDC and DSM-IV-defined MEL patients had significantly higher basal cortisol levels than NMD patients, while DSM-III and DSM-III-R-defined subtypes did not.
Veltman et al. 2018	192 (M = 138)	M = 70.5 (7.1) N = 68.3 (6.7)	M = 68.8 N = 83.3	Latent class analysis	IDS M = 32.6 (14.2) N = 32.2 (11.8)	Salivary cortisol wakening curves, μg/dL Area under the curve with respect to ground (AUCg) and increase (AUCi), diurnal slope (mean decline per hour)	AUCg N > M AUCi M > N Slope M > N	N = atypical depression. Described as severe subtypes. Self-calculated Welch’s tests all insignificant.
Hypothalamic-pituitary-adrenal axis processes
Amsterdam et al. 1989 (Study 1)	72 (M = 51)	M = 38.33 (12.71) N = 34.43 (9.74)	M = 66.67 N = 71.43	DSM-III	Not reported	Post-infusion serum cortisol levels (μg/dL) CC = cumulative cortisol (area under hormone response curve) after ACTH infusion MC = maximum increase after ACTH infusion PC = peak levels after cosyntropin infusion	CC M > N MC M > N PC M > N	Infusion with ACTH or cosyntropin, a derivative of ACTH. Self-calculated Welch’s test. Mimics processes of HPA axis.
Amsterdam et al. 1989 (Study 2)	26 (M = 14)	35.7 (9.6)	Not reported	DSM-III	Not reported	Post-infusion plasma ACTH (pg/mL) and cortisol levels (μg/mL) CC = cumulative hormone (area under hormone response curve) after CRH infusion MC = maximum increase after CRH infusion PC = peak levels after CRH infusion	ACTH CC N > M MC N > M PC N > M Cortisol CC M > N MC M > N PC M > N	None significant according to self-calculated Welch’s tests. However, follows expected directions. Mimics processes of HPA axis.
Amsterdam et al. 1989 (Study 5)	32 (M = 22)	39.13 (13.04)	71.88	DSM-III	Not reported	Serum cortisol levels (μg/dL) Cumulative cortisol (area under hormone response curve) after CRH infusion (C-CRH) Maximum response to ACTH infusion (M-ACTH)	C-CRH M > N M-ACTH M > N	C-CRH significant and M-ACTH trend for significance from self-calculated Welch test. Replicated elements of previous studies, but in new sample before they began treatment.
Maes et al. 1990	96 (M = 44)	48.1 (14.8)	100	Cluster analysis of DSM-III symptoms	HDRS-17 21.6 (5.3)	Post-DST ACTH, pg/mL	M > N	Looks at ‘vital’ (melancholic) and ‘non-vital’ (non-melancholic) classes. Reported ANOVA test. Age and HDRS scores reported for entire sample of 100; 4 excluded.
Maes, Meltzer, et al. 1994	45 (M = 21)	M = 51.7 (9.62) N = 45.3 (13.23)	Not reported	DSM-III (SCID)	HDRS-17 M = 26.2 (2.75) N = 21.8 (2.45)	Plasma ACTH, pg/mL Basal, post-DST (PDST), post-DST & CRH (PDC), change in AUC of concentration-time curve (AUC), peak minus post-DST (P-PD), peak minus basal (P-MB)	All M > N Basal, PDST	See Tables 3 and 4 in study. Basal ACTH, post-DST ACTH, and post-DST β-endorphin significantly higher in MEL compared to NMD patients according to self-calculated Welch’s test.
Cizza et al. 2012	42 (M = 32)	M = 34.3 (7.0) N = 36.6 (6.8)	100	DSM-IV (SCID)	HDRS M = 8.1 (5.3) N = 10.6 (8.1)	Plasma ACTH (pg/mL)	N > M	All N are atypical. Age and HDRS reported for whole sample (n = 89, 51 MEL). Non-significance reported.
Thyroid hormones
Thyrotropin-releasing hormone (TRH) Challenge Test
Hubain et al. 1994	280 M = 57	44 (13.4)	66.1	NEDDI	HDRS-24 Female = 27 (7) Male = 28 (8)	TRH test max change in serum TSH (μIU/mL)	N > M	M = 6.2 (4), N = 8 (5). After TRH challenge, the maximum observed increases in serum TSH were lower in Newcastle-defined endogenous and psychotic depressed patients than for NMD patients.
Rush et al. 1997	133 (M = 85)	M = 39.2 (13.1) N = 36.2 (10.2)	M = 65.9 N = 60.4	RDC (SADS-L)	HDRS-17 M = 28.3 (8.5) N = 18.7 (4.1)	Blunted post-test serum TRH rates, μIU/mL Blunting cut-offs of 4 and 5μIU/mL given	All M > N	Blunted response in 24.7%/21.2% (5/4 μIU/mL) of endogenous and 10.4%/8.3% of non-endogenous depressed patients. Self-calculated χ² tests showed trends for significance differences.
Natural hormone levels
Brouwer et al. 2005	113 M = 32 N1 = 25 N2 = 56	47 (11)	62	DSM-IV (SCID)	HDRS-17 21 (3)	TSH (mU/L), FT4 (pmol/L)	TSH N1 > M N2 > M FT4 M > N1 N2 > M	N1 = atypical depression, N2 = unspecified depression. Reported p values showed no differences in FT4 or TSH.
Gjerris et al. 1985	33 (M = 15)	Not reported	Not reported	NEDDI	BRMS, but not reported	TRH in cerebrospinal fluid (ng/L)	M > N.	Medians (range): M = 6.1 (2.7–10.5), N = 5.6 (2.4–11.6). Non-significance reported.
Hubain et al. 1994	280 M = 57	44 (13.4)	66.1	NEDDI	HDRS-24 Female = 27 (7) Male = 28 (8)	Serum TSH (μIU/mL)	See notes	Reported insignificant. But basal TSH negatively correlated with Newcastle scores, suggesting lower levels in endogenous depression.
Maes et al. 1990	96 (M = 44)	48.1 (14.8)	100	DSM-III-R (SCID) Cluster analysis of DSM-III symptoms	HDRS-17 21.6 (5.3)	Basal TSH (μIU/mL) and FT4 (pmol/L), unspecified transformation	TSH N > M FT4 M > N	See Table 13 in paper. Looks at ‘vital’ (melancholic) and ‘non-vital’ (non-melancholic) classes. Reported ANOVA test. Age and HDRS scores reported for entire sample of 100; 4 excluded.
Maes, Maes, et al. 1992	80 M = 35	M = 53.1 (13.02) N = 41.4 (12.75)	0	Cluster analysis	HDRS M = 26.9 (5.32) N = 18.7 (4.02)	Plasma basal TSH and FT4, units not reported	TSH N > M FT4 N > M	Reported ANOVA tests. Replicated Maes et al. (1990), but the FT4 result was in the opposite direction and slightly over the threshold for significance.
Maes, Meltzer, et al. 1993	158 M = 57	M = 54.2 (12.8) N = 45.8 (13.1)	M = 72.5 N = 81.2	DSM-III-R (SCID)	HDRS-17 M = 27.3 (4.5) N = 21.8 (3.0)	Basal TSH (μIU/mL), FT4 (pmol/L), FT3 (pmol/L),	TSH N > M FT4 M > N FT3 M > N	Self-calculated Welch’s tests. Used a TSH assay test purported to be more accurate than the TRH test. 8.8% of their melancholic sample displayed subclinical hypothyroidism.
Türkçapar et al. 1999	65 RDC M = 43 DSM-III M = 16 DSM III-R M = 32 DSM-IV M = 37 ICD-10 M = 41	Overall 33.3 (9.9) RDC M = 35.2 (9.8) N = 29.6 (9.3) DSM-III M = 35 (11.1) N = 32.6 (10.8) DSM III-R M = 37 (10.3) N = 28 (7.3) DSM-IV M = 35.3 (9.9) N = 30.7 (9.4) ICD-10 M = 35.6 (10.3) N = 29.4 (7.9)	63.1	RDC DSM-III (SCID) DSM III-R DSM-IV ICD-10	HDRS-17 Overall 24.0 (6.3) RDC M = 26.2 (6.1) N = 20.0 (4.5) DSM-III M = 26.8 (5.2) N = 23.2 (6.4) DSM III-R M = 25 (6.7) N = 23.1 (5.9) DSM-IV M = 26.2 (6.8) N = 21.3 (4.4) ICD-10 M = 25.9 (6.6) N = 21.0 (4.4)	Plasma TSH, μIU hTSH/mL	RDC N > M DSM-III N > M DSM III-R N > M DSM-IV N > M ICD-10 N > M	See Tables 2 and 3 in study. DSM-III-R and DSM-IV-defined MEL patients showed significantly lower levels of TSH than NMD patients, while the DSM-III and RDC-defined subtypes did not.
Other hormones
Gjerris et al. 1985	ICD-9 46 (M = 32) NEDDI 37 (M = 14)	ICD-9 Median M = 52 Median N = 36	ICD-9 M = 52.4 N = 69.2	ICD-9 NEDDI	BRMS, but not reported	Vasopressin in cerebrospinal fluid (ng/L)	M > N	Medians (range): M = 1.2 (1.0-2.2), N = 1.2 (0.8-2.1). Non-significance reported.
Maes, Meltzer, et al. 1994	45 (M = 21)	M = 51.7 (9.62) N = 45.3 (13.23)	Not reported	DSM-III (SCID)	HDRS-17 M = 26.2 (2.75) N = 21.8 (2.45)	Plasma β-endorphin pg/mL	M > N	See Table 4 in study. Basal β-endorphin non-significant according to self-calculated Welch test.

Acc: accuracy; BDI: Beck Depression Inventory; BRMS: Bech-Rafaelsen Melancholia Scale; DSM: Diagnostic and Statistical Manual for Mental Disorders; HDRS: Hamilton Depression Rating Scale; ICD: International Classification of Diseases; IDS: Inventory of Depressive Symptomatology; M: melancholic class; MA: melancholia with delusions and agitation; MADRS: Montgomery-Asberg Depression Rating Scale; MD: melancholia with delusions; MDA: melancholia with delusions and agitation; MINI: Mini International Neuropsychiatric Interview; N: non-melancholic depression class; NEDDI: Newcastle Endogenous Depression Diagnostic Index; NPV: negative predictive value; PPV: positive predictive value; RDC: Research Diagnostic Criteria; SADS: Schedule for Affective Disorders and Schizophrenia; SCID: Structural Clinical Interview for DSM; TNR: true negative rate; TPR: true positive rate. Bold indicates a statistically significant difference.

Table 4. Summaries of studies concerning neurological biomarkers of melancholia.

Study	Sample size	Age (mean (SD))	Sex (% female)	MEL diagnostic criteria	Depression severity (mean (SD))	Measurement method	Summary
Brain morphology
Korgaonkar et al. 2011	29 (M = 11)	M = 43.1 (15.3) N = 38.9 (16.3)	M = 63.6 N = 55.6	CORE	HDRS M = 20.1 (3.2) N = 18.4 (2.7)	Volumetric MRI and DTI (global GM, WM, CSF) WM (frontal, parietal, occipital, temporal) DTI clusters across limbic-DLPFC-thalamic pathway	White matter integrity was compromised for MEL patients in areas associated with the limbic system, dorsolateral prefrontal cortex (PFC), thalamus projections, and the corpus callosum. However, significant differences were only found between MEL patients and controls, and not between MEL and NMD patients. Author hypothesise that this pattern may be due to unique features within MEL subtype of patients, that result in either decreased myelination or altered WM microstructure
Pizzagalli et al. 2004	38 (M = 20)	M = 36.5 (12.9) N = 33.1 (8.8)	M = 65 N = 55.6	DSM-IV (SCID)	BDI M = 33.8 (6.4) N = 29.5 (8.1) HDRS M = 18.2 (5.1) N = 20.1 (5.5)	Structural MRI	Considered structural brain differences in the subgenual PFC with regards to grey (GM) and white matter (WM) as measured by homogeneity probability maps, which model the probability that the brain area being studied belongs to a particular tissue type. MEL patients demonstrated reduced subgenual PFC activity, including increased delta current density (16.9%) and lower glucose metabolism (16.4%) compared to NMEL groups. No differences between MEL and NMD patients were found on GM or WM probabilities, though the GM probabilities of MEL patients did appear to be negatively correlated with age (i.e. older subjects, lower GM density in subgenual PFC). However, authors noted insufficient granularity to detect whether subtler structural variations were present. Wide-ranging study that also had implications for EEG activity and metabolic markers (see below).
Exner et al. 2009	35 (M = 26)	M = 35.0 (10.5) N = 33.0 (10.0)	M = 76.9 N = 44.4	DSM-IV (SCID)	HDRS-21 M = 22.1 (4.4) N = 16.3 (4.5)	Anterior supplementary motor area volume (pre-SMA; left and right), total brain volume, both mL	pre-SMA-L: M = 10.8 (3.1), N = 10.2 (4.2) pre-SMA-R: 10.3 (2.2), N = 10.9 (4.8) Total: M = 1207 (130), N = 1233 (127) Reported post-hoc ANOVA tests: MEL patients have significantly larger left pre-SMA and significantly smaller right pre-SMA. Total brain volume not significant.
Mertse et al. 2022	54 (M = 35)	M = 45.0 (12) N = 41.9 (12)	M = 45.8 N = 47.4	HES	HDRS-21 M = 23.7 (5) N = 18.3 (5) BDI M = 29.6 (9) N = 26.8 (8)	Caudal ACC thickness (left and right, mm)	Left M = 2.019 (0.2876), N = 2.278 (0.2980), right M = 2.123 (0.175), N = 2.256 (0.2065). Reported ANCOVA contrasts after controlling for age and gender. For both sides, MEL patients have thinner caudal ACC.
Takahashi et al. 2016	29 (M = 10) (N includes 3 atypical)	32.5 (8.3)	75.86	DSM-IV-TR (SCID)	BDI 36.8 (8.9)	Olfactory sulcus morphology	Reported no differences.
Takahashi et al. 2020	29 (M = 10) (N includes 3 atypical)	32.5 (8.3)	75.86	DSM-IV-TR (SCID)	BDI 36.8 (8.9)	Volume of pineal gland (total and parenchymal, mm^3)	Parenchymal M = 142.0 (40.7), N = 101.9 (42.4). Volume not given for total volume, but both volumes are reported significantly higher in MEL patients. Used same sample as Takahashi et al. (2016).
Brain networks
Hyett, Breakspear et al. 2015	32 (M = 16)	M = 38.00 (9.94) N = 40.44 (10.73)	M = 50.0 N = 62.5	Custom criteria	QIDS-16 M = 16.69 (4.22) N = 15.06 (4.10)	fMRI data: graph theoretic measures of brain connectivity (in-degree; out-degree; edge and regional connectivity)	The first in a series of three functional resonance magnetic imaging (fMRI) studies by Hyett et al. MEL compared to NMD patients showed strong disconnection in multiple brain networks related to attention and interoception. These were found in attentional networks leading from the anterior insula, and connections between networks of the right frontoparietal and insula regions.
Hyett, Parker, et al. 2015	32 (M = 16)	M = 38.00 (9.94) N = 40.44 (10.73)	M = 50.0 N = 62.5	Custom criteria	QIDS-16 M = 16.69 (4.22) N = 15.06 (4.10)	fMRI data: connectivity amongst brain nodes during positive and negative film viewing dynamic causal modelling of naturalistic film viewing	Follow up study using the same sample as Hyett, Breakspear, et al. (2015). Increases in connectivity were observed when engaging with negatively salient film viewing. The cortical networks of attention and interception displayed increased connectivity unique to MEL patients when shifting from rest to negative stimuli.
Hyett et al. 2018	45 (M = 22)	M = 43.50 (13.17) N = 39.45 (12.71)	M = 42.9 N = 57.1	Custom criteria	QIDS-SR-16 M = 16.77 (3.41) N = 14.91 (4.50)	Diffusion magnetic resonance imaging (dMRI): voxel-wise fractional anisotropy (FA), radial diffusivity (RD), axial diffusivity (AD)	Evidence for disturbance in white matter in the internal capsule for those with melancholia. Specifically, there is increased FA in right retrolenticular limb, and further differences when compared with controls.
Guo et al. 2016	34 (M = 17)	M = 42.06 (11.81) N = 40.44 (10.73)	M = 52.9 N = 52.9	Custom criteria	QIDS-SR-16 M = 16.82 (4.19) N = 15.06 (4.11)	Statistical analyses of whole-brain functional imaging data	Follow up study of the two Hyett papers above. Examined brain activity during the viewing of emotionally salient films (both positive and negative). MEL patients showed considerably reduced activity in the default mode network, less connectivity in multiple hub regions, and particularly decreased connectivity in the subgenual anterior cingulate cortex (ACC) The latter particularly contrasted with increased connectivity that was observed in NMD patients.
Regonia et al. 2021	120 (M = 89)	M = 43.37 (11.65) N = 41.48 (9.46)	M = 49.44 N = 51.61	DSM-IV (MINI)	BDI-II M = 32.40 (7.88) N = 29.13 (6.08)	Energy landscape analysis of resting state fMRI data	Considerable differences in basin energy, basin frequency, and transition dynamics in networks associated with the basal ganglia, the default mode network for dorsal regions, and left regions of executive control networks.
Workman et al. 2016	63 (M = 33)	M = 37.7 (11.1) N = 35 (13.4)	M = 72.7 N = 56.7	DSM-IV-TR (SCID)	BDI M = 3.9(4.1) N = 3.2(4.2) MADRS M = 37.3 (5.1) N = 32.8 (5.9)	Resting-state fMRI: disconnection in remitted patients	MEL and NMD patients can be differentiated due to the former displaying less (and abnormal) connectivity between the subgenual cingulate cortex and two brain regions. These were the left amygdala and the right parahippocampal gyrus.
Zhang et al. 2021	59 (M = 31)	M = 28.65 (5.30) N = 32.04 (8.18)	M = 67.7 N = 64.3	DSM-IV	HDRS-17 M = 21.77 (3.79) N = 21.00 (3.14)	Differences in fALFF values of global brain regions	Studied the fractional amplitude of low-frequency fluctuations (fALFF), a proxy measure of spontaneous intrinsic brain activity. MEL patients were found to have notably lower fALFF values than NMD patients in the left anterior cingulate cortex. Same sample used for machine learning studies below.
Machine learning classification
Shan et al. 2021	59 (M = 31)	M = 28.65 (5.30) N = 32.04 (8.18)	M = 67.7 N = 64.3	DSM-IV	HDRS-17 M = 21.77 (3.79) N = 21.00 (3.14)	fMRI to review voxel-mirrored homotopic connectivity (VMHC) and pattern classification	The first in a series of three fMRI studies by Guo, Xie, et al. that used machine learning (and specifically support vector machines) to differentiate MEL from NMD patients on the basis of regional network-related brain activity.  MEL and NMD patients were compared in their observed patterns of voxel-mirrored homotopic connectivity (VHMC), a method of examining interhemispheric resting-state functional connectivity. Compared to NMD patients, MEL patients displayed reduced VHMC in the precentral and postcentral gyri. SVMs could classify MEL and NMD patients on the basis of precentral gyrus VHMC with an accuracy of 74.58%.
Yan, He, et al. 2021	59 (M = 31)	M = 28.65 (5.30) N = 32.04 (8.18)	M = 67.7 N = 64.3	DSM-IV	HDRS-17 M = 21.77 (3.79) N = 21.00 (3.14)	Neuroimaging (VBM with ReHO analyses)	Examined regional homogeneity (ReHo), a method for examining time-related changes in blood oxygen level-dependent (BOLD) signals. Compared to NMD patients, MEL patients displayed lower ReHo in the middle and right superior occipital gyri. When considered with abnormal ReHo in the right fusiform gyrus and cerebellum, SVMs could distinguish between MEL and NMD patients with 83.05% accuracy.
Yan, Cui, et al. 2021	59 (M = 31)	M = 28.65 (5.30) N = 32.04 (8.18)	M = 67.7 N = 64.3	DSM-IV	HDRS-17 M = 21.77 (3.79) N = 21.00 (3.14)	Neuroimaging (fMRI)	Examined network homogeneity (NH) in the default mode network. Compared to NMD patients, MEL patients displayed increased NH in the superior medial frontal gyrus, and decreased NH in the left inferior temporal gyrus. SVM models examining NH in these regions resulted in a diagnostic accuracy of 79.66%. They also reported positive correlations between NH in MEL patients (and not NMD patients) and anhedonia, and particularly anhedonia specific to wanting particular events to happen.
Electroencephalographic (EEG) activity
Pizzagalli et al. 2002	38 (M = 20)	M = 36.5 (12.9) N = 33.1 (8.8)	M = 65 N = 55.6	DSM-IV (SCID)	BDI M = 33.8 (6.4) N = 29.5 (8.1) HDRS M = 18.2 (5.1) N = 20.1 (5.5)	Brain electrical tomography (Whole Brain LORETA)	Same sample as Pizzagalli et al. (2004). MEL compared to NMD patients showed particularly strong beta3 activity (21.5–30.0 Hz) in the inferior and right superior frontal lobes. Specifically, these were in Brodmann areas (BAs) 9-11.
Pizzagalli et al. 2004	38 (M = 20)	M = 36.5 (12.9) N = 33.1 (8.8)	M = 65 N = 55.6	DSM-IV (SCID)	BDI M = 33.8 (6.4) N = 29.5 (8.1) HDRS M = 18.2 (5.1) N = 20.1 (5.5)	Brain electrical tomography (whole brain LORETA)	MEL and NMD patients showed differences in electroencephalogram (EEG) readings in the delta (1.5–6.0 Hz) and alpha1 (8.5–10.0 Hz) bands in multiple brain regions. However, the most notable findings were found in four specific clusters. Compared to NMD patients, MEL patients showed delta activity that was higher in the subgenual PFC and lower in the precuneus, while alpha1 activity was also showed to be higher in the left middle temporal gyrus and lower in the right superior temporal gyrus. Once age and gender were controlled for, only delta subgenual PFC activity remained significantly different between MEL and NMD patients.
Nieber and Schlegel 1992	63	53.8 (18.1)	65.08	BRMS	BRMS 19.2 (6.9)	EEG activity	Greater melancholic psychomotor retardation correlated with more slow EEG activity (i.e. theta2, alpha1). Was also correlated with less fast activity (i.e. alpha3, beta).
Quinn et al. 2014	117 (M = 57)	Not reported	Not reported	DSM-IV (MINI) CORE	Not reported	EEG alpha symmetry	NMD patients showed greater left frontal activation than MEL patients, potentially reflecting NMD’s greater proneness to anger and irritability.
Neural correlates of behaviour/cognition
Tsujii et al. 2014	60 (M = 32)	M = 40.8 (15.3) N = 38.9 (11.8)	M = 50 N = 46.4	DSM-IV (MINI)	HDRS-17 M = 15.4 (5.2) N = 13.2 (5.4) BDI-II M = 35.4 (12.2) N = 31.4 (10.9)	Haemodynamic activity on a verbal fluency task (VFT) as measured by multichannel near-infrared spectroscopy (NIRS)	The VFT induced significantly smaller changes in mean oxygenated haemoglobin across multiple points of the right temporal region for MEL compared to NMD patients. Differing patterns of correlations between brain activity in the two groups and psychomotor retardation were also observed. Specifically, MEL patients showed a positive correlation in the right temporal regions. Meanwhile, NMD patients showed a negative correlation in the frontal and left temporal regions.
Tsujii et al. 2016	82 (M = 30)	M = 42.2 (11.8) N = 40.6 (11.7)	M = 50 N = 65.4	DSM-IV (MINI)	HDRS-17 M = 15.1 (4.8) N = 13.1 (3.6)	Haemodynamic activity on a verbal fluency task (VFT) as measured by multichannel near-infrared spectroscopy (NIRS)	Follow up study of Tsujii et al. (2014) using a similar methodology. The time before the task until the task onset was typified by increased activity in frontotemporal regions for NMD but not for MEL patients. Supports previous point of Tsujii et al. (2014). During the task MEL patients had significantly less activity than NMD patients in multiple areas covering, amongst others, the bilateral temporal region and dorsolateral and ventrolateral PFCs.
Foti et al. 2014	34 (M = 11) (N includes 6 atypical)	26.00 (8.89)	100	DSM-IV-TR (SCID)	Not reported	Multimodal neuroimaging, response to reward	Depressed patients with impaired mood reactivity (a key feature of melancholic depression) showed a blunted ventral striatum (VS) response to reward (in this case, within the context of a laboratory gambling task) in comparison to those with non-impaired mood reactivity. However, VS activation was not related to melancholia as defined by DSM criteria. Once anhedonia and depressive symptom severity was controlled for, impaired mood reactivity was no longer associated with VS activity. Also examined sensory evoked potentials (see below).
Shankman et al. 2011	65 (M = 17)	M = 35.5 N = 33.2	M = 52.9 N = 70.8	DSM-IV (SCID), confirmed with HES	HDRS M = 30.6 (6.9) N = 24.5 (7.8)	EEG asymmetry, response to reward	EEG asymmetry during reward processing in posterior brain regions (i.e. asymmetrical in the left and right posterior regions) could be a biomarker of melancholic depression. This was as a result of their observation of different asymmetries between MEL and NMD patients, specifically driven by comparatively lower right and left posterior activity in MEL and NMD patients respectively.
Liu et al. 2016	141 (M = 38)	M = 35.92 (12.86) N = 32.79 (11.66)	M = 60.53 N = 68.93	DSM-IV (SCID) HES	HDRS M = 29.00 (8.12) N = 25.35 (7.38)	EEG asymmetry, response to reward	EEG data suggested that abnormal processing of reward is typical of melancholia, which remained significant even when accounting for severity of both overall depression and NMD. However, this is only the case when it was defined on the basis of a dimensional scale (HES) rather than a categorical entity (which, like Foti et al. (2014), was defined on the basis of DSM criteria).
Miscellaneous metabolic markers
Pizzagalli et al. 2004	38 (M = 20)	M = 36.5 (12.9) N = 33.1 (8.8)	M = 65 N = 55.6	DSM-IV (SCID)	BDI M = 33.8 (6.4) N = 29.5 (8.1) HDRS M = 18.2 (5.1) N = 20.1 (5.5)	Glucose metabolism (PET)	MEL patients had lower glucose metabolism than NMD patients in the subgenual PFC, though there were no significant differences with controls. They also found higher glucose metabolism in those with melancholia in multiple other brain areas (BAs 2, 6, 9, 22-24, 37, 40, 42, 46). However, they noted that these could only be considered as preliminary results due to the low sample size.
Biver et al. 1994	12 (M = 6)	37.6 (13.2)	50	NEDDI	HDRS-24 31 (7)	Glucose metabolism (PET)	Glucose metabolism in the parietal and frontal lobes was examined in endogenous and non-endogenous depressed patients. No differences were found.
Austin et al. 1992	40	45.8 (13.7)	55	NEDDI	HDRS-17 22.4 (5.6)	Regional cerebral blood flow through single photon emission computed tomography (SPECT) using 99mTc-exametazine (99TE), neuropsychological testing	Endogenous depression could be characterised by a greater uptake of 99TE in the cingulate and frontal cortices. 99TE is a compound used for photon emission tomography, which is metabolised in a proportional manner to cerebral blood flow and thus can be considered a proxy for brain activity.
Fountoulakis et al. 2004	50	41.0 (11.4)	70	DSM-IV/ICD-10 (SCAN 2.0)	HDRS, but not reported	Regional cerebral blood flow through single photon emission computed tomography (SPECT) using HMPAO.	Examined regional cerebral blood flow through the use of SPECT. Atypical depression was characterised by greater perfusion in the right frontal lobe and less in the right occipital lobe. Melancholia was characterised by perfusion levels in the opposite direction.
Sensory-evoked potentials (behavioural)
Elton 1984	28 (M = 7)	M = 33 N = 34	N = 57.1	ICD-8 DSM-III	BDI, but not reported	Event-related potentials, reaction to stimuli	P240 peak latencies were shorter in endogenous depression compared to reactive depression.
Foti et al. 2014	34 (M = 11) (N includes 6 atypical)	26.00 (8.89)	100	DSM-IV-TR (SCID)	MASQ but not reported	Multimodal neuroimaging, feedback-related negativity (FRN; the difference wave of the neural responses to reward and punishment in a behavioural task)	No differences were found for DSM-defined melancholia, but depressed patients with impaired mood reactivity showed a more blunted FRN amplitude than those with regular mood reactivity, which was attributed to a lesser response to reward rather than punishment. This association remained even after adjusting for anhedonia and symptom severity. Differences in response to reward and punishment were observed in those with regular mood reactivity, but not those with impaired mood reactivity (the latter being a common symptom of melancholia).
Quinn, Harris, et al. 2012	114 (M = 60)	M = 41.58 (16.06) N = 38.08 (12.32)	M = 63.3 N = 42.6	DSM-IV (MINI)	HDRS M = 21.77 (4.71) N = 19.39 (5.12) DASS-D M = 28.87 (8.82) N = 26.89 (8.82)	EEG, ERP, activity on a Go/No-Go task	MEL compared to NMD patients showed less inhibitory control. However, this was not reflected in any differences in event-related potentials.
Weinberg et al. 2016	50 (M = 17)	M = 22.83 (3.28) N = 23.15 (3.01)	M = 66.67 N = 73.53	DSM-IV (SCID)	Not reported	Error related negativity (ERNs)	In response to errors on a behavioural task, the magnitude of the neural response (ERN) was less in remitted MEL compared to remitted NMD patients.
Auditory-evoked potentials
Fitzgerald et al. 2009	27 (M = 14)	M = 49.07 (12.11) N = 38.46 (14.11)	M = 66.67 N = 92.31	DSM-IV-TR (MINI)	MADRS M = 32.57 (7.92) N = 19.23 (8.96) BDI-II M = 32.79 (5.56) N = 24.46 (10.96)	Intensity independence of auditory evoked potential (IDEAP)	The slope of the IDEAP was considerably shallower for MEL compared to NMD patients.
Kerr et al. 2011	83 (M = 49)	M = 38 (13) N = 35 (10)	M = 67.3 N = 50	DSM-IV (MINI)	HDRS M = 21.2 (6.4) N = 18.8 (3.9) DASS-D M = 29.5 (10.4) N = 24.8 (9.9)	EEG audio event-related potentials measured by deconvolution (a wave superposition technique that reduces redundancies and elucidates previously unobtainable waveform properties) and thalamocortical modelling (which describes overall neural properties of region-based subpopulations defined by a theoretical model of neuronal dynamics).	Deconvolution established that D2 latency in prefrontal areas was higher for MEL patients. Thalamocortical modelling reported that MEL patients had longer signal propagation times, and a thalamocortical gain (i.e. the change in firing rate across thalamic or cortical connections that are caused by a firing rate change in other components of the chain) that was 40% less excitatory.
Schlegel et al. 1991	36	44 (13)	61.1	BRMS	HDRS, but not reported	P300 latency, auditory event-related potentials	Depressed patients who scored higher on the BRMS also tended to have higher P300 latencies. This was apparent for both the BRMS total score and on items related to psychomotor retardation.
Kemp et al. 2010	105 (M = 57)	M = 37.63 (12.88) N = 34.66 (9.74)	M = 66.67 N = 52.08	DSM-IV (MINI)	HDRS M = 22.03 (4.67) N = 18.04 (4.40) DASS-D M = 31.71 (7.86) N = 21.67 (8.73)	Event-related potentials associated with auditory oddball selective attention task	M patients showed an increased P200 response to target and non-target stimuli, decreased P300 response to targets, and slowed latency in the N200/P300 complex.
Visual-evoked potentials and ocular features
Fotiou et al. 2003	50 = MEL	M = 47.00 (13.03) N = 37.00 (7.79)	70	DSM-IV/ICD-10 (SCAN 2.0)	HDRS-17 M = 28.53 (6.54) N = 23.14 (3.82) HDRS-21 M = 30.05 (7.43) N = 25.57 (4.33)	Electroretinographic and flash-visual evoked potential (VEP) recording	Melancholic compared to atypically depressed patients had greater P100 and N80 latencies.
Winograd-Gurvich et al. 2006a	19 (M = 9)	M = 47.78 N = 40.80	M = 77.8 N = 60	CORE	MADRS M = 28.78 N = 27.40	Oculomotor differences (i.e. reflexive vs visually guided saccades)	The first of two studies that examined MEL/NMD differences on ocular motor tasks. MEL compared to NMD patients had higher saccadic latencies and decreased saccadic peak velocities (SPVs) in response to targets with larger positional amplitudes (SPVs were increased for NMD patients).
Winograd-Gurvich et al. 2006b	19 (M = 9)	M = 47.78 N = 40.80	M = 77.8 N = 60	CORE	MADRS M = 28.78 N = 27.40	Self-paced and reprogrammed saccades in relation to behavioural tasks relying on fronto-striatal brain regions.	Follow up of Winograd-Gurvich et al. (2006a). Compared to NMD patients, MEL patients were less accurate and had (again) lower saccadic peak velocities.
Fountoulakis et al. 2005	30 (M = 16)	M = 47.00 (13.03) N = 37.00 (7.79)	Not reported for sub-sample	DSM-IV (SCAN 2.0)	HDRS-17 M = 28.53 (6.54) N = 23.14 (3.82)	Data from flash electroretinogram (flash-ERG) and electrooculogram (EOG)	No reported differences for EOG measures. Flash-ERG showed no differences either, but melancholic anhedonia was associated with larger b wave latency. N = atypical depression.
Sleep
Giles et al. 1986	103 (M = 46)	35.1 (9.9)	Not reported	RDC	HDRS-17 23.2 (5.2)	REM	Symptoms of with melancholic depression were associated with reduced REM latencies. Specifically, these were anhedonia and a non-reactive mood.
Giles et al. 1990	177 (M = 75)	Age groups 20–34 = 92 35–49 = 57 50–72 = 28	68.93	RDC (SADS-L)	HDRS-17 age groups 20–34 = 21.8 (5.5) 35–49 = 21.9 (5.3) 50–72 = 23.9 (5.5)	REM latency values	Showed the same effect as Giles et al. (1986) in patients with formal RDC diagnoses of endogenous and non-endogenous depression. Older age also appeared to be associated with greater REM latency for non-endogenously depressed patients only.
Hubain et al. 1995	39 (M = 15)	M = 38 (5) N = 33 (7)	0	NEDDI	HDRS-24 M = 30 (11) N = 24.7 (5)	Sleep EEG parameters (i.e. sleep period time (SPT), REM latency, stage II, slow wave sleep (SWS), REM latency)	Endogenously depressed patients more frequently showed a REM latency exceeding 50 min. Also examined patients with regards to sleep time and slow wave sleep but did not detect any differences between endogenous and non-endogenous depressed patients.
Rush et al. 1982	70 (M = 32)	36.5 (11.3)	Roughly two thirds	RDC (SADS-L)	HDRS-17 M = 26.7 (5.1) N = 21.4 (5.8) BDI M = 30.6 (8.3) N = 27.5 (9.5)	REM latency	A REM latency of 62 min provided reasonable differentiability between endogenous and non-endogenous depression. Sensitivity 66%, specificity 79%.
Rush et al. 1997	133 (M = 85)	M = 39.2 (13.1) N = 36.2 (10.2)	M = 65.9 N = 60.4	RDC (SADS-L)	HDRS-17 M = 28.3 (8.5) N = 18.7 (4.1)	REM, measured with EEG, EOG, and EMG	REM latency in 65 and 34% of endogenous and non-endogenous patients respectively.
Riemann et al. 1994	178 (M = 113)	M = 43.6 (13.2) N = 37.1 (10.8)	M = 67.3 N = 66.2	DSM-III	HDRS-21 M = 27.3 (5.5) N = 24.3 (5.7)	Sleep parameters (continuity, architecture, REM sleep)	Did not find any differences in both baseline and (cholinergic) induced REM sleep.
Hein et al. 2019	16 (M = 8)	M = 41 (12) N = 48 (14) Median (IQR)	Not reported	NEDDI	HDRS-24 M = 23.5 (12) N = 20 (9)	Sleep parameters (REM and slow-wave sleep)	Compared to patients with non-endogenous (neurotic) depression, endogenously depressed patients showed more evidence for brain network activity corresponding to a so-called ‘small-world organisation’ during REM sleep. Small-world organisation is typified by high clustering and short path lengths.
Brain-derived neurotrophic factor (BDNF)
Patas et al. 2014	1037 (M = 161)	M = 41.9 (11.6) N = 40.8 (12.2)	M = 61.5 N = 69.2	Custom algorithm	IDS-30, but not reported	Serum levels, ng/mL	M = 8.9 (3.5), N = 9.0 (3.4). No differences in BDNF levels. However, higher BDNF could be predicted by higher IL-6 levels in MEL but not NMD subjects.
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	M = 26.5 (7.4), N = 25.8 (7.5). No differences.
Neurotransmitters
Lehto et al. 2006	29 (M = 10) (N1 = 8) (N2 = 11)	M = 28.10 (8.14) N1 = 32.38 (11.45) N2 = 25.45 (5.84)	M = 100 N1 = 75 N2 = 72.7	DSM-IV (SCID)	HDRS-21 M = 26.80 (6.49) N1 = 19.13 (3.40) N2 = 18.36 (4.50) HDRS-29 M = 32.80 (8.88) N1 = 27.25 (5.50) N2 = 28.64 (9.60)	Densities of serotonin (SERT) and dopamine transporter (DAT) in the midbrain	Reported no differences in SERT or DAT between any of the diagnostic subgroups. N1 = non-differentiated, N2 = atypical depression.

BDI: Beck Depression Inventory; BRMS: Bech-Rafaelsen Melancholia Scale; CSF: cerebrospinal fluid; DASS-D: Depression Anxiety and Stress Scale (Depression Subscale); DSM: Diagnostic and Statistical Manual for Mental Disorders; EEG: electroencephalogram; fMRI: functional magnetic resonance imaging; GM: grey matter; HDRS: Hamilton Depression Rating Scale; HES: Hamilton Endogenomorphic Scale; ICD: International Classification of Diseases; IDS: Inventory of Depressive Symptomatology; LORETA: Low-resolution brain electromagnetic tomography; M: melancholic class; MADRS: Montgomery-Asberg Depression Rating Scale; MINI: Mini International Neuropsychiatric Interview; N: non-melancholic depression class; NEDDI: Newcastle Endogenous Depression Diagnostic Index; PET: positron emission tomography; QIDS: Quick Inventory of Depressive Symptoms; RDC: Research Diagnostic Criteria; SADS: Schedule for Affective Disorders and Schizophrenia; SCAN: Schedules for Clinical Assessment in Neuropsychiatry; SCID: Structural Clinical Interview for DSM; WM: white matter

Table 5. Summaries of studies exploring immunological markers of melancholia.

Study	Sample size	Age (mean (SD))	Sex (% female)	MEL diagnostic criteria	Depression severity (mean (SD))	Measurement method	Results	Notes
C-reactive protein (CRP)
Glaus et al. 2014	660 (M = 198)	Not reported	M = 66.67 N = 63.42	DSM-IV (DIGS)	Not reported	Serum levels, pg/mL	See notes.	Only regression coefficients given, βM = −0.06.
Karlović et al. 2012	55 (M = 32)	M = 48.6 (8.1) N = 50.9 (8.3)	M = 37.5 N = 34.8	DSM-IV-TR (SCID)	HDRS-17 M = 30.1 (10.7) N = 21.5 (7.9)	Serum levels, mg/L	M > N	M = 3.22 (3.0), N = 2.23 (2.0). N = atypical depression
Lamers et al. 2013	233 (M = 111)	M = 40.2 (12.1) N = 39.6 (12.1)	M = 65.8 N = 79.5	Latent class and latent transition analysis	IDS-30 M = 39.1 (10.0) N = 35.2 (11.5)	Log transformed serum levels, mg/L	N > M	M = 1.18 (3.57), N = 1.86 (3.48). N = atypical depression Described as severe subtypes.
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	See notes	Regression model with multiple biomarkers, β = 0.169, p = 0.069. N = atypical depression Specified as ‘severe’ melancholia/atypical depression.
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	Serum levels of high sensitivity CRP, mg/L	N1 > M M = N2	M = 1.00 (0.50, 3.20), N1 = 1.20 (0.60, 2.60), N2 = 1.00 (0.50, 2.10). All medians with interquartile ranges. βM = 0.05. N1 = atypical depression, N2 = unspecified depression.
Liu et al. 2022	81 (M = 44)	M = 27.8 (7.2) N = 26.9 (7.1)	M = 65.9 N = 56.8	DSM-5 (confirmed with MINI)	HDRS M = 23.3 (4.3) N = 20.4 (3.1) IDS-30 M = 47.3 (10.5) N = 40.2 (9.3) QIDS-SR M = 21.9 (6.3) N = 18.9 (5.5)	Log (base 2) transformed serum levels of high sensitivity CRP	N > M	N = atypical depression
Mohamed et al. 2020	52 (M = 19) (N1 = 11) (N2 = 22)	M = 36.5 (10.4) N1 = 35.7 (11.06) N2 = 35.4 (12.66)	M = 52.6 N1 = 90.9 N2 = 68.2	DSM-5 (SCID)	Not reported	Serum levels of high sensitivity CRP, mg/L	N > M	M = 3.24 (3.61), N = 5.76 (5.27). High sensitivity CRP. Combined NMD. N1 = atypical depression N2 = unspecified depression
Rothermundt, Arolt, Peters, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS Overall = 27 (7) M = 29 (7) N = 25 (6)	Serum levels, mg/dL	M > N	M = 0.25 (0.13), N = 0.19 (0.14)
Rudaz et al. 2017	172 (M = 43) (N1 = 31) (N2 = 98)	M = 47.2 (7.3) N1 = 51.0 (7.9) N2 = 49.0 (7.8)	M = 65.1 N1 = 67.7 N2 = 52.0	DSM-IV (DIGS)	Not reported	Log transformed serum levels of high sensitivity CRP	M > N1 M > N2	M = 1.1 (0.5, 2.5), N1 = 0.7 (0.4, 2.1), N2 = 0.8 (0.4, 1.6). All medians and interquartile ranges. N1 = atypical depression, N2 = unspecified depression.
Sánchez-Carro et al. 2023	91 (M = 65)	Overall = 50.64 (10.18) M = 50.41 (9.96) N = 51.19 (10.89)	M = 69.23 N = 76.92	DSM-IV-TR (MINI)	Not reported	Log transformed serum levels, mg/dL	M > N	M = −0.76 (0.39), N = −0.86 (0.34)
Veltman et al. 2018	192 (M = 138)	M = 70.5 (7.1) N = 68.3 (6.7)	M = 68.8 N = 83.3	Latent class analysis	IDS M = 32.6 (14.2) N = 32.2 (11.8)	High-sensitivity plasma levels	N > M	M = 2.1 (1.8), N = 2.7 (2.1). Specified as being ‘severe’ melancholic and atypical depression.
Vogelzangs et al. 2014	358 (M = 50) (N1 = 23) (N2 = 285)	70.6 (7.3)	66.2	Latent class analysis	IDS M = 47.6 (7.1) N1 = 41.4 (8.7) N2 = 26.5 (11.0)	Plasma levels, mg/L	See notes	Only regression coefficients given. N2 reference. βM = −0.06, βN2 = 0.06. N1 = atypical depression. N2 = generic non-melancholic depression.
Tumour Necrosis Factor alpha (TNFα)
Dunjic-Kostic et al. 2013	47 (M = 29)	M = 50.28 (7.41) N = 52.26 (7.29)	M = 55.17 N = 55.56	DSM-IV	HDRS M = 25.83 (0.51) N = 22.41 (0.79)	Serum levels, pg/mL	M > N	M = 7.46 (5.85), N = 5.67 (2.91). N = atypical depression
Glaus et al. 2014	660 (M = 198)	Not reported	M = 66.67 N = 63.42	DSM-IV (DIGS)	Not reported	Serum levels, pg/mL	See notes	Only regression coefficients given. βM = −0.06.
Huang and Lee 2007	42 (M = 25)	M = 40.2 (8.0) N = 34.7 (7.6)	71.4	DSM-IV (SCID)	HDRS M = 35.8 (4.0) N = 33.5 (4.3)	Serum levels, pg/mL	M > N	M = 1.1 (2.6), N = 0.7 (2.1). ANCOVA controlling for age, BMI
Karlović et al. 2012	55 (M = 32)	M = 48.6 (8.1) N = 50.9 (8.3)	M = 37.5 N = 34.8	DSM-IV-TR (SCID)	HDRS-17 M = 30.1 (10.7) N = 21.5 (7.9)	Serum levels, pg/mL	M > N	M = 6.91 (2.9), N = 5.86 (1.8). N = atypical depression
Lamers et al. 2013	233 (M = 111)	M = 40.2 (12.1) N = 39.6 (12.1)	M = 65.8 N = 79.5	Latent class and latent transition analysis	IDS-30 M = 39.1 (10.0) N = 35.2 (11.5)	Log transformed serum levels, pg/mL	N > M	M = 0.78 (1.89), N = 1.03 (1.97). N = atypical depression. Described as severe subtypes.
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	Serum levels, pg/mL	N1 > M M > N2	M = 2.72 (1.62, 4.29), N1 = 2.96 (1.84, 4.93), N2 = 2.61 (1.58, 4.18). All medians and interquartile ranges. βM = 0.02. N1 = atypical depression, N2 = unspecified depression.
Liu et al. 2022	81 (M = 44)	M = 27.8 (7.2) N = 26.9 (7.1)	M = 65.9 N = 56.8	DSM-5 (confirmed with MINI)	HDRS M = 23.3 (4.3) N = 20.4 (3.1) IDS-30 M = 47.3 (10.5) N = 40.2 (9.3) QIDS-SR M = 21.9 (6.3) N = 18.9 (5.5)	Log (base 2) transformed serum levels	M > N	N = atypical depression
Maes, Mihaylova, et al. 2012	37 (M = 12)	42.0 (11.0)	73.5	DSM-IV-TR (SCID)	HDRS 22.6 (3.1)	Serum levels, pg/mL	M > N	M = 14.76 (4.24), N = 10.59 (3.91). HDRS was significantly related to the TNFα levels (r = 0.36, p = 0.03).
Patas et al. 2014	1037 (M = 161)	M = 41.9 (11.6) N = 40.8 (12.2)	M = 61.5 N = 69.2	Custom algorithm	IDS-30, but not reported	Plasma levels, pg/mL	M = N	M = 0.80 (0.60, 1.0), N = 0.80 (0.60, 1.10). Medians with interquartile ranges.
Rudaz et al. 2017	172 (M = 43) (N1 = 31) (N2 = 98)	M = 47.2 (7.3) N1 = 51.0 (7.9) N2 = 49.0 (7.8)	M = 65.1 N1 = 67.7 N2 = 52.0	DSM-IV (DIGS)	Not reported	Log transformed serum levels, pg/mL	M > N1 N2 > M	M = 2.4 (1.8, 4.9), N1 = 2.0 (1.5, 4.8), N2 = 2.6 (1.5, 4.0). All medians with interquartile ranges. N1 = atypical depression, N2 = unspecified depression
Sánchez-Carro et al. 2023	91 (M = 65)	Overall = 50.64 (10.18) M = 50.41 (9.96) N = 51.19 (10.89)	M = 69.23 N = 76.92	DSM-IV-TR (MINI)	Not reported	Log transformed serum levels, pg/mL	N > M	M = 0.78 (0.23), N = 0.83 (0.31).
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	M > N	M = 1.02 (0.35), N = 0.97 (0.29). Medians and interquartile ranges.
Interleukin 6 (IL-6)
Dunjic-Kostic et al. 2013	47 (M = 29)	M = 50.28 (7.41) N = 52.26 (7.29)	M = 55.17 N = 55.56	DSM-IV	HDRS M = 25.83 (0.51) N = 22.41 (0.79)	Serum levels, pg/mL	M > N	M = 12.54 (14.68), N = 7.06 (5.08). N = atypical depression
Glaus et al. 2014	660 (M = 198)	Not reported	M = 66.67 N = 63.42	DSM-IV (DIGS)	Not reported	Serum levels, pg/mL	See notes	Only regression coefficients given, βM = 0.10.
Karlović et al. 2012	55 (M = 32)	M = 48.6 (8.1) N = 50.9 (8.3)	M = 37.5 N = 34.8	DSM-IV-TR (SCID)	HDRS-17 M = 30.1 (10.7) N = 21.5 (7.9)	Serum levels, pg/mL	M > N	M = 3.11 (1.6), N = 2.45 (1.2). N = atypical depression
Lamers et al. 2013	233 (M = 111)	M = 40.2 (12.1) N = 39.6 (12.1)	M = 65.8 N = 79.5	DSM-IV for MDD (CIDI) Latent class and latent transition analysis	IDS-30 M = 39.1 (10.0) N = 35.2 (11.5)	Log transformed serum levels, pg/mL	N > M	M = 0.75 (2.64), N = 1.04 (2.42). N = atypical depression. Described as severe subtypes.
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	Serum levels, pg/mL	N1 > M M > N2	M = 1.22 (0.52, 2.81), N1 = 1.29 (0.56, 3.47), N2 = 1.19 (0.50, 3.21). All medians and interquartile ranges. βM = 0.08. N1 = atypical depression, N2 = unspecified depression
Liu et al. 2022	81 (M = 44)	M = 27.8 (7.2) N = 26.9 (7.1)	M = 65.9 N = 56.8	DSM-5 (confirmed with MINI)	HDRS M = 23.3 (4.3) N = 20.4 (3.1) IDS-30 M = 47.3 (10.5) N = 40.2 (9.3) QIDS-SR M = 21.9 (6.3) N = 18.9 (5.5)	Log (base 2) transformed serum levels	M > N	N = atypical depression
Maes, Scharpé, et al. 1993	17 (M = 10)	53.5 (14.6)	Not reported	DSM-III (SCID)	HDRS-17 M = 21.6 (3.8) N = 27.5 (3.2)	Production, ng/mL	N > M	M = 29.9 (18.3), N = 56.0 (31.5)
Marques-Deak et al. 2007	46 (M = 28)	39.4 (10.2)	100	DSM-IV (SCID)	HDRS 34.5 (7.8)	Serum levels, ng/mL	M > N	M = 136.2 (100.0), N = 126.5 (48.7).
Patas et al. 2014	1037 (M = 161)	M = 41.9 (11.6) N = 40.8 (12.2)	M = 61.5 N = 69.2	Custom algorithm	IDS-30, but not reported	Plasma levels, pg/mL	M > N	M = 0.93 (0.61, 1.46), N = 0.77 (0.49, 1.28). Medians with interquartile ranges.
Rudaz et al. 2017	172 (M = 43) (N1 = 31) (N2 = 98)	M = 47.2 (7.3) N1 = 51.0 (7.9) N2 = 49.0 (7.8)	M = 65.1 N1 = 67.7 N2 = 52.0	DSM-IV (DIGS)	Not reported	Log transformed serum levels	N1 > M M > N2	M = 1.5 (0.7, 2.6), N1 = 1.7 (0.4, 3.3), N2 = 1.2 (0.6, 3.4). Medians with interquartile ranges. N1 = atypical depression, N2 = unspecified depression
Veltman et al. 2018	192 (M = 138)	M = 70.5 (7.1) N = 68.3 (6.7)	M = 68.8 N = 83.3	Latent class analysis	IDS M = 32.6 (14.2) N = 32.2 (11.8)	Plasma levels	N > M	M = 0.47 (1.32), N = 0.65 (1.80). Medians with interquartile ranges. Specified as being ‘severe’ melancholic and atypical depression.
Vogelzangs et al. 2014	358 (M = 50) (N1 = 23) (N2 = 285)	70.6 (7.3)	66.2	Latent class analysis	IDS M = 47.6 (7.1) N1 = 41.4 (8.7) N2 = 26.5 (11.0)	Plasma levels, mg/L	See notes	Only regression coefficients given. N2 reference. βM = −0.03, βN2 = −0.02. N1 = atypical depression. N2 = generic non-melancholic depression.
Interleukin 1 (α and β) (IL-1α/β)
Glaus et al. 2014	660 (M = 198)	Not reported	M = 66.67 N = 63.42	DSM-IV (DIGS)	Not reported	Serum levels of IL-1β, pg/mL	See notes.	Only odds ratios given. OR = 0.99
Huang and Lee 2007	42 (M = 25)	M = 40.2 (8.0) N = 34.7 (7.6)	71.4	DSM-IV (SCID)	HDRS M = 35.8 (4.0) N = 33.5 (4.3)	Serum levels of IL-1β	M > N	M = 5.8 (8.6), N = 1.9 (3.7). ANCOVA controlling for age, BMI
Kaestner et al. 2005	37 (M = 21)	M = 51.4 (12.0) N = 36.8 (12.4)	M = 71.4 N = 68.8	DSM-IV (SCID) NEDDI	MADRS M = 34 (6) N = 28 (7)	Serum levels of IL-1β, pg/mL	N > M
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	Serum levels of IL-1β, pg/mL	M > N1 M > N2	M = 0.43 (0.10, 1.98), N1 = 0.31 (0.10, 1.66), N2 = 0.42 (0.10, 2.00). Medians with interquartile ranges. ORM = 0.91. N1 = atypical depression, N2 = unspecified depression.
Maes et al. 2012	37 (M = 12)	42.0 (11.0)	73.5	DSM-IV-TR (SCID)	HDRS 22.6 (3.1)	Serum levels of IL-1 (i.e. α + β), pg/mL	M > N	M = 6.74 (3.05), N = 5.11 (2.67). HDRS was significantly related to the TNFα levels (r = 0.36, p = 0.03).
Marques-Deak et al. 2007	46 (M = 28)	39.4 (10.2)	100	DSM-IV (SCID)	HDRS 34.5 (7.8)	Serum levels of IL-1β, ng/mL	N > M	M = 33.8 (18.5), N = 40.4 (18.4).
Rothermundt, Arolt, Peters, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS Overall = 27 (7) M = 29 (7) N = 25 (6)	Serum levels of IL-1β, pg/mL	N > M
Rudaz et al. 2017	172 (M = 43) (N1 = 31) (N2 = 98)	M = 47.2 (7.3) N1 = 51.0 (7.9) N2 = 49.0 (7.8)	M = 65.1 N1 = 67.7 N2 = 52.0	DSM-IV (DIGS)	Not reported	Log transformed serum levels of IL-1β, pg/mL	M > N1 M = N2	M = 0.5 (0.1, 2.9), N1 = 0.3 (0.1, 1.7), N2 = 0.5 (0.1, 1.5). Medians with interquartile ranges. N1 = atypical depression, N2 = unspecified depression
Sowa-Kućma et al. 2018	114 (M = 37)	49.38 (10.79)	64.04	DSM-IV-TR (SCID)	HDRS 12.97 (9.02) MADRS 17.49 (13.45)	Z scores of log transformed serum levels of IL-1α, pg/mL	N > M	M = −0.37 (0.88), N = 0.13 (1.09), Regression OR = 0.40. Regression model controlled for soluble interleukin-6 receptor and two principal components derived from medication usage data.
Interferon gamma (IFNγ)
Marques-Deak et al. 2007	46 (M = 28)	39.4 (10.2)	100	DSM-IV (SCID)	HDRS 34.5 (7.8)	Serum levels, IU/mL	N > M	M = 189.0 (259.0), N = 210.5 (184.9)
Rothermundt, Arolt, Fenker, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS M = 28.73 (7.03) N = 24.91 (6.40)	Production, pg/mL	N > M	Reported Mann–Whitney U test.
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	N > M	M = 1.45 (0.72), N = 1.64 (0.72). Medians with interquartile ranges.
Haptoglobin
Erdem et al. 2011	82 (M = 37)	M = 35.6 (6.2) N = 34.0 (6.6)	M = 37.8 N = 46.7	DSM-IV (SCID)	HDRS M = 30.1 (6.0) N = 23.9 (4.2)	Serum levels, nmol/L	M > N	M = 159.2 (71.0), N = 95.8 (45.7)
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	DSM-IV (CIDI) for MDD Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	See notes	Regression model with multiple biomarkers. t = 0.051, p = 0.460. N = atypical depression. Specified as ‘severe’ melancholia/atypical depression.
Rothermundt, Arolt, Peters, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS Overall = 27 (7) M = 29 (7) N = 25 (6)	Serum levels, mg/dL	M > N	M = 116.9 (39.8), N = 110 (53.2)
Interleukin 2 (IL-2)
Rothermundt, Arolt, Fenker, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS M = 28.73 (7.03) N = 24.91 (6.40)	Production, pg/mL	N > M	Reported Mann–Whitney U test.
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	N > M	M = 0.24 (0.07), N = 0.27 (0.12). Medians with interquartile ranges.
Interleukin 10 (IL-10)
Rothermundt, Arolt, Fenker, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS M = 28.73 (7.03) N = 24.91 (6.40)	Production, pg/mL	N > M	M = 155.2 (143.5), N = 271.6 (205.6)
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	M > N	M = 0.34 (0.25), N = 0.28 (0.52). Medians with interquartile ranges.
Receptors
Sowa-Kućma et al. 2018	114 (M = 37)	49.38 (10.79)	64.04	DSM-IV-TR (SCID)	HDRS 12.97 (9.02) MADRS 17.49 (13.45)	Z scores of log transformed serum levels, pg/mL (sIL-2R), ng/Ml (sIL-1RA, sIL-6R, sTNF-R1, sTNF-R2)	See notes (sIL-6R)	sIL-6R, M = 0.47 (0.94), N = −0.17 (1.09), Regression OR = 29.27. Regression model covariates included IL-1α, sIL-6R, and two principal components derived from medication usage data. sIL-1RA, sIL-2R, sTNF-R1, and sTNF-R2
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels (IL-1RA, IL-2Rβ, IL-6R, TNF-R1, TNF-R2)	See notes	Regression model with multiple biomarkers, none significant. N = atypical depression. Specified as ‘severe’ melancholia/atypical depression.
α2-Macroglobulin
Rothermundt, Arolt, Peters, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS Overall = 27 (7) M = 29 (7) N = 25 (6)	Serum levels, g/L	M > N	M = 190.6 (56.1), N = 161.6 (52.7)
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	See notes	Regression model with multiple biomarkers, β = −0.013. N = atypical depression Specified as ‘severe’ melancholia/atypical depression.
Interleukin 4 (IL-4)
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	M > N	M = 0.81 (0.37), N = 0.75 (0.20). Medians with interquartile ranges.
Interleukin 17 (IL-17)
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels	M > N	M = 7.99 (10.02), N = 7.51 (9.7). Medians with interquartile ranges.
Leukocytes and associated proteins
Maes, Lambrechts, et al. 1992	24 (M = 12)	M = 49.9 (15.2) N = 41.9 (13.9)	Not reported	DSM-III (but also DSM-III-R SCID)	HDRS M = 26.7 (4.2) N = 21.2 (3.1)	Counts (WBC × 10^-3), cells/mm^3 white blood cells (WBC), lymphocytes, monocytes, granulocytes, pan-T cells, HLA-DR antigen complex, IL-2 receptor, pan B cells, surface Ig, T-helper cells, T-suppressor cells, T-helper/suppressor ratio, T-memory cells, T-virgin Cells	See notes (T-suppressor cells, T-helper/suppressor ratio)	T-suppressor cells: M = 545 (197.5), N = 357 (239.0). T-helper/suppressor ratio, M = 1.63 (0.83), N = 2.43 (1.14). Rest are insignificant.
Rothermundt, Arolt, Fenker, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS M = 28.73 (7.03) N = 24.91 (6.40)	Absolute count (NK cells, lymphocytes)	NK cells N > M Lymphocytes N > M	NK cells: M = 469.9 (310.6), N = 629.4 (458.7). Lymphocytes given in graph.
Rothermundt, Arolt, Peters, et al. 2001	43 (M = 22)	44.45 (9.95)	65.12	DSM-IV, ICD-10 (CIDI)	HDRS Overall = 27 (7) M = 29 (7) N = 25 (6)	Count, /µl	N > M	M = 408 (121), N = 564 (222)
Neopterin
Maes, Mihaylova, et al. 2012	37 (M = 12)	42.0 (11.0)	73.5	DSM-IV-TR (SCID)	HDRS 22.6 (3.1)	Serum levels, ng/mL	M > N	M = 5.54 (2.22), N = 2.31 (1.26). HDRS was significantly related to the neopterin levels (r = 0.56, p < 0.001).
Maes, Scharpé, et al. 1994	31 (M = 18)	M = 52.7 (12.1) N = 50.8 (10.1)	M = 77.78 N = 76.92	DSM-III (SCID)	HDRS M = 25.1 (4.2) N = 18.4 (2.7)	Log-transformed plasma levels, nmol/L	M > N	M = 7.7 (7.6), N = 6.5 (3.6). No significant difference from self-calculated Welch’s test.
Neutrophil gelatinase-associated lipocalin (NGAL)
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	See notes	Regression model with multiple biomarkers. β = −0.009. N = atypical depression. Specified as ‘severe’ melancholia/atypical depression.
Veltman et al. 2018	192 (M = 138)	M = 70.5 (7.1) N = 68.3 (6.7)	M = 68.8 N = 83.3	Latent class analysis	IDS M = 32.6 (14.2) N = 32.2 (11.8)	Log transformed plasma levels	M > N	M = 66.7 (37.4), N = 57.5 (20.6). Specified as being ‘severe’ melancholic and atypical depression.
5-HT antibody positivity
Maes, Ringel, et al. 2012	109 (M = 35)	43.0 (11.0)	54.1	DSM-IV-TR	HDRS 22.59 (3.06)	Positivity rates	M > N	M = 82.9%, N = 40.5%. Significant according to self-calculated χ² test.

BDI: Beck Depression Inventory; CIDI: Composite International Diagnostic Interview; DIGS: Diagnostic Interview for Genetic Studies; DSM: Diagnostic and Statistical Manual for Mental Disorders; HDRS: Hamilton Depression Rating Scale; HES: Hamilton Endogenomorphic Scale; ICD: International Classification of Diseases; IDS: Inventory of Depressive Symptomatology; M: melancholic class; MADRS: Montgomery-Asberg Depression Rating Scale; MINI: Mini International Neuropsychiatric Interview; N: non-melancholic depression class; NEDDI: Newcastle Endogenous Depression Diagnostic Index; QIDS: Quick Inventory of Depressive Symptoms; RDC: Research Diagnostic Criteria; SADS: Schedule for Affective Disorders and Schizophrenia; SCAN: Schedules for Clinical Assessment in Neuropsychiatry; SCID: Structural Clinical Interview for DSM. Bold indicates a statistically significant difference.

Table 6. Summaries of studies exploring metabolic markers of melancholia.

Study	Sample size	Age (mean (SD))	Sex (% female)	MEL diagnostic criteria	Depression severity (mean (SD))	Measurement method	Results	Notes
Lipid levels
Cizza et al. 2012	42 (M = 32)	M = 34.3 (7.0) N = 36.6 (6.8)	100	DSM-IV (SCID)	HDRS M = 8.1 (5.3) N = 10.6 (8.1)	mg/dL, triglycerides log-transformed HDL, LDL, TC, TG	HDL: M > N Rest: N > M	Non-melancholic class included 16 patients with atypical depression.
Eriksson et al. 2023	392 (M = 119)	M = 61 (3) N = 62 (3)	M = 51.3 N = 73.3	DSM-IV (BDI)	BDI M = 14.1 (5.2) N = 14.3 (5.1)	mmol/L HDL, LDL, TC, TG	HDL N > M LDL N > M TC N > M TG M > N	LDL: M = 3.40 (0.82), N = 3.65 (0.92). TC M = 5.71 (1.26), N = 5.99 (1.09).
Huang 2005	109 (M = 39)	M = 31.6 (8.7) N = 31.3 (8.5)	M = 82.1 N = 64.3	DSM-IV		mg/dL (non-ratio measures only) HDL, LDL, VLDL, TC, TG, TC:HDL ratio, LDL:HDL ratio	HDL: M > N Rest: N > M	N = atypical depression.
Lamers et al. 2013	233 (M = 111)	M = 40.2 (12.1) N = 39.6 (12.1)	M = 65.8 N = 79.5	Latent class and latent transition analysis	IDS-30 M = 39.1 (10.0) N = 35.2 (11.5)	mmol/L, TG was log transformed HDL, TG	HDL M > N TG N > M	HDL: M = 1.63 (0.42), N = 1.52 (0.44). TG: M = 1.06, N = 1.26. N = atypical depression. Described as severe subtypes. Means adjusted for sex, age, educational level, and smoking status.
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	HDL, mmol/L	M = N1 M = N2	M = 1.7 (0.4), N1 = 1.7 (0.5), N2 = 1.7 (0.4). N1 = atypical depression, N2 = unspecified depression
Sánchez-Carro et al. 2023	91 (M = 65)	Overall = 50.64 (10.18) M = 50.41 (9.96) N = 51.19 (10.89)	M = 69.23 N = 76.92	DSM-IV-TR (MINI)	Not reported	Log transformed HDL, TG	HDL M > N TG N > M	HDL: M = 1.78 (0.14), N = 1.76 (0.13). TG: M = 1.98 (0.20), N = 2.01 (0.19)
Selenius et al. 2021	296 (M = 97)	M = 61.6 (3.0) N = 62.2 (3.3)	M = 52.6 N = 75.9	DSM-IV (BDI)	Not reported	mmol/L HDL, TC, TG	TC N > M Rest M = N
Vogelzangs et al. 2014	358 (M = 50) (N1 = 23) (N2 = 285)	70.6 (7.3)	66.2	Latent class analysis	IDS M = 47.6 (7.1) N1 = 41.4 (8.7) N2 = 26.5 (11.0)	HDL, mmol/L	See notes	Only regression coefficients given. N2 reference. βM = 0.01, βN2 = −0.05. N1 = atypical depression. N2 = generic non-melancholic depression.
Oxidative stress
Sánchez-Carro et al. 2023	91 (M = 65)	Overall = 50.64 (10.18) M = 50.41 (9.96) N = 51.19 (10.89)	M = 69.23 N = 76.92	DSM-IV-TR (MINI)	Not reported	Log transformed HNE, GSSG/GSH	HNE N > M GSSG/GSH N > M	GSSG/GSH: M = 0.28 (0.54), N = 0.55 (0.57)
Sowa-Kućma et al. 2018	114 (M = 37)	49.38 (10.79)	64.04	DSM-IV-TR (SCID)	HDRS 12.97 (9.02) MADRS 17.49 (13.45)	Z scores of log transformed TBARS levels, nmol/mL	N > m	M = 0.17 (1.01), N = 0.29 (0.91)
Spanemberg et al. 2014	33 (M = 13)	M = 52.8 (10.7) N = 48.4 (7.7)	M = 76.9 N = 90.0	CORE	HDRS M = 22.6 (2.9) N = 21.8 (2.9)	Serum levels of TBARS and PCC	TBARS N > M PCC M = N	TBARS: M = 8.93 (2.66), N = 10.93 (5.66). PCC: M = 0.06 (0.08), N = 0.06 (0.03). All medians and interquartile ranges.
Leptin
Cizza et al. 2012	42 (M = 32)	M = 34.3 (7.0) N = 36.6 (6.8)	100	DSM-IV (SCID)	HDRS M = 8.1 (5.3) N = 10.6 (8.1)	Log-transformed 24-hr plasma levels, ng/mL	M = N	N = atypical depression.
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	See notes (significant)	Regression model with multiple biomarkers. β = 0.227. N = atypical depression Specified as ‘severe’ melancholia/atypical depression.
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	ng/mL	N1 > M M > N2 Regression for N1	M = 10.86 (12.19), N1 = 11.78 (14.83), N2 = 9.97 (13.21). All medians with interquartile ranges. βM = 0.06, βN1 = 0.16, βN2 = 0.01. N1 = atypical depression, N2 = unspecified depression
Tryptophan
Maes et al. 1990	96 (M = 44)	48.1 (14.8)	100	Cluster analysis of DSM-III symptoms	HDRS-17 21.6 (5.3)	× 10^-6mol/L (TRP) TRP, TRP:CAA ratio	All N > M	TRP: M = 46 (10), N = 57 (3). TRP:CAA: M = 7.55 (1.39), N = 9.39 (1.67). Looks at ‘vital’ (melancholic) and ‘non-vital’ (non-melancholic) classes. Reported ANOVA test. Age and HDRS scores reported for entire sample of 100; 4 excluded.
Maes, Scharpé, et al. 1994	31 (M = 18)	M = 52.7 (12.1) N = 50.8 (10.1)	M = 77.78 N = 76.92	DSM-III (SCID)	HDRS M = 25.1 (4.2) N = 18.4 (2.7)	Square root (TRP) or log-transformed (TRP:CAA) plasma levels, μmol/L (TRP)	All M > N	TRP: M = 59.6 (9.97), N = 57.7 (14.2). TRP:CAA: M = 0.090 (0.017), N = 0.088 (0.016)
Adiponectin
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	See notes	Regression model with multiple biomarkers. β = −0.051. N = atypical depression. Specified as ‘severe’ melancholia/atypical depression.
Lasserre et al. 2017	1233 (M = 369) (N1 = 179) (N2 = 685)	M = 49.5 (8.7) N1 = 49.1 (8.9) N2 = 49.4 (8.8)	M = 68.3 N1 = 72.1 N2 = 62.8	DSM-IV	Not reported	μg/mL	N1 > M M > N2	M = 9.33 (8.33), N1 = 9.36 (7.29), N2 = 8.79 (8.11). Medians with interquartile ranges. βM = −0.01, βN1 = −0.03, βN2 = 0.02. N1 = atypical depression, N2 = unspecified depression
Angiotensin-converting enzyme (ACE)
Lamers et al. 2016	359 (M = 231)	M = 41.7 (12.0) N = 40.7 (11.7)	M = 68.0 N = 71.9	Latent class analysis	IDS-30 M = 38.6 (9.7) N = 39.1 (8.8)	Log (base 10) transformed serum levels	Negative association with M	Regression model with multiple biomarkers. β = 0.059, p = 0.001. N = atypical depression. Specified as ‘severe’ melancholia/atypical depression.
Maes, Scharpé, et al. 1992	25 (M = 13)	M = 47.7 (12.97) N = 54.7 (11.09)	Not reported	DSM-III	HDRS M = 26.2 (3.61) N = 21.3 (3.12)	Units per litre (u/L)	N > M	M = 242.9 (109.2), N = 421.3 (232.4). Examined the impact upon the HPA axis of angiotensin-I-converting enzyme (ACE), an enzyme involved in facilitating the creation of hormones that increase catecholamine release, which have been suggested to induce ACTH secretion.
Zinc
Maes, D'Haese, et al. 1994	32 (M = 18)	M = 52.7 (12.1) N = 51.0 (9.7)	Not reported	DSM-III (SCID)	HDRS M = 25.1 (4.2) N = 18.4 (2.6)	Serum levels, mg/L	M > N	M = 1.78 (0.22), N = 1.77 (0.25). No significant difference from self-calculated Welch’s test.
Styczeń et al. 2017	114	49.4 (10.7)	75.44	DSM-IV-TR	MADRS, but not reported	Serum levels, mg/L	N > M	M = 0.92 (0.18), N = 1.00 (0.40). No significant difference from reported Mann–Whitney U test.

BDI: Beck Depression Inventory; CIDI: Composite International Diagnostic Interview; DIGS: Diagnostic Interview for Genetic Studies; DSM: Diagnostic and Statistical Manual for Mental Disorders; HDL: high-density cholesterol; HDRS: Hamilton Depression Rating Scale; ICD: International Classification of Diseases; IDS: Inventory of Depressive Symptomatology; LDL: low-density cholesterol; M: melancholic class; MADRS: Montgomery-Asberg Depression Rating Scale; MINI: Mini International Neuropsychiatric Interview; N: non-melancholic depression class; NEDDI: Newcastle Endogenous Depression Diagnostic Index; RDC: Research Diagnostic Criteria; SADS: Schedule for Affective Disorders and Schizophrenia; SCID: Structural Clinical Interview for DSM; TC: total cholesterol; TG: triglycerides; VLDL: very low-density cholesterol. Bold indicates a statistically significant difference.

Figure 2. Strategy for future research into creating a biologically-informed nosology for melancholia.

Carney MWP, Roth M, Garside RF. 1965. The diagnosis of depressive syndromes and the prediction of E.C.T. response. Br J Psychiatry. 111(477):659–674.

 PubMedGoogle Scholar

Spitzer RL, Endicott J, Robins E. 1978. Research diagnostic criteria: rationale and reliability. Arch Gen Psychiatry. 35(6):773–782.

 PubMed Web of Science ®Google Scholar

Parker G, Hadzi-Pavlovic D, Boyce P, Wilhelm K, Brodaty H, Mitchell P, Hickie I, Eyers K. 1990. Classifying depression by mental state signs. Br J Psychiatry. 157(1):55–65.

 PubMedGoogle Scholar

Parker G, Hadzi-Pavlovic D, Wilhelm K, Hickie I, Brodaty H, Boyce P, Mitchell P, Eyers K. 1994. Defining melancholia: properties of a refined sign-based measure. Br J Psychiatry. 164(3):316–326.

 PubMedGoogle Scholar

Feinberg M, Carroll BJ. 1982. Separation of subtypes of depression using discriminant analysis: i. separation of unipolar endogenous depression from non-endogenous depression. Br J Psychiatry. 140(4):384–391.

 PubMedGoogle Scholar

Brown RP, Stoll PM, Stokes PE, Frances A, Sweeney J, Kocsis JH, Mann JJ. 1988. Adrenocortical hyperactivity in depression: effects of agitation, delusions, melancholia, and other illness variables. Psychiatry Res. 23(2):167–178.

 PubMed Web of Science ®Google Scholar

Carroll BJ, Feinberg M, Greden JF, Tarika J, Albala AA, Haskett RF, James NM, Kronfol Z, Lohr N, Steiner M, et al. 1981. A specific laboratory test for the diagnosis of melancholia: standardization, validation, and clinical utility. Arch Gen Psychiatry. 38(1):15–22.

 PubMed Web of Science ®Google Scholar

Davidson J, Lipper S, Zung WW, Strickland R, Krishnan R, Mahorney S. 1984. Validation of four definitions of melancholia by the dexamethasone suppression test. Am J Psychiatry. 141(10):1220–1223.

 PubMed Web of Science ®Google Scholar

Georgotas A, Stokes PE, Hapworth WE, Kim OM, Fanelli C, Stoll PM, Sinaiko E, McCue RE. 1986. The relationship of the dexamethasone suppression test to subtypes of depression and to symptomatic severity in the elderly. J Affect Disord. 10(1):51–57.

 PubMed Web of Science ®Google Scholar

Holden NL. 1983. Depression and the Newcastle Scale: their relationship to the dexamethasone suppression test. Br J Psychiatry. 142(5):505–507.

 PubMedGoogle Scholar

Miller KB, Nelson JC. 1987. Does the dexamethasone suppression test relate to subtypes, factors, symptoms, or severity? Arch Gen Psychiatry. 44(9):769–774.

 PubMed Web of Science ®Google Scholar

Rush AJ, Giles DE, Roffwarg HP, Parker CR. 1982. Sleep EEG and dexamethasone suppression test findings in outpatients with unipolar major depressive disorders. Biol Psychiatry. 17(3):327–341.

 PubMed Web of Science ®Google Scholar

Amsterdam JD, 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.

 PubMed Web of Science ®Google Scholar

Brown WA, Shuey I. 1980. Response to dexamethasone and subtype of depression. Arch Gen Psychiatry. 37(7):747–751.

 PubMed Web of Science ®Google Scholar

Evans DL, Nemeroff CB. 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 Psychiatr Res. 21(2):185–194.

 PubMed Web of Science ®Google Scholar

Peselow ED, Fieve RR. 1992. Depressive attributional style and the dexamethasone suppression test: relationship to the endogenous/melancholic distinction and to each other. Psychopathology. 25(4):173–182.

 PubMed Web of Science ®Google Scholar

Rush AJ, Giles DE, Schlesser MA, Orsulak PJ, Parker CR, Weissenburger JE, Crowley GT, Khatami M, Vasavada N. 1996. The dexamethasone suppression test in patients with mood disorders. J Clin Psychiatry. 57(10):470–484.

 PubMed Web of Science ®Google Scholar

Rush AJ, Giles DE, Schlesser MA, Orsulak PJ, Weissenburger JE, Fulton CL, Fairchild CJ, Roffwarg HP. 1997. Dexamethasone response, thyrotropin-releasing hormone stimulation, rapid eye movement latency, and subtypes of depression. Biol Psychiatry. 41(9):915–928.

 PubMed Web of Science ®Google Scholar

Hubain PP, Staner L, Dramaix M, Kerkhofs M, Papadimitriou G, Mendlewicz J, Linkowski P. 1998. The dexamethasone suppression test and sleep electroencephalogram in nonbipolar major depressed inpatients: a multivariate analysis. Biol Psychiatry. 43(3):220–229.

 PubMed Web of Science ®Google Scholar

Maes M, Lambrechts J, Bosmans E, Jacobs J, Suy E, Vandervorst C, De Jonckheere C, Minner B, Raus J. 1992. Evidence for a systemic immune activation during depression: results of leukocyte enumeration by flow cytometry in conjunction with monoclonal antibody staining. Psychol Med. 22(1):45–53.

 PubMed Web of Science ®Google Scholar

Maes M, Meltzer H, Cosyns P, Calabrese J, D'Hondt P, Blockx P. 1994. Adrenocorticotropic hormone, β-endorphin and cortisol responses to oCRH in unipolar depressed patients pretreated with dexamethasone. Prog Neuropsychopharmacol Biol Psychiatry. 18(8):1273–1292.

 PubMed Web of Science ®Google Scholar

Maes M, Scharpé S, Meltzer HY, Bosmans E, Suy E, Calabrese J, Cosyns P. 1993. Relationships between interleukin-6 activity, acute phase proteins, and function of the hypothalamic-pituitary-adrenal axis in severe depression. Psychiatry Res. 49(1):11–27.

 PubMed Web of Science ®Google Scholar

Türkçapar MH, Akdemir A, Örsel SD, Demirergi N, Sirin A, Kiliç EZ, Özbay MH. 1999. The validity of diagnosis of melancholic depression according to different diagnostic systems. J Affect Disord. 54(1–2):101–107.

 PubMed Web of Science ®Google Scholar

Brouwer JP, Appelhof BC, Hoogendijk WJG, Huyser J, Endert E, Zuketto C, Schene AH, Tijssen JGP, Van Dyck R, Wiersinga WM, et al. 2005. Thyroid and adrenal axis in major depression: a controlled study in outpatients. Eur J Endocrinol. 152(2):185–191.

 PubMed Web of Science ®Google Scholar

Cizza G, Ronsaville DS, Kleitz H, Eskandari F, Mistry S, Torvik S, Sonbolian N, Reynolds JC, Blackman MR, Gold PW, 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):e28912.

 PubMed Web of Science ®Google Scholar

Karlović D, Serretti A, Vrkić N, Martinac M, Marčinko D. 2012. Serum concentrations of CRP, IL-6, TNF-α and cortisol in major depressive disorder with melancholic or atypical features. Psychiatry Res. 198(1):74–80.

 PubMed Web of Science ®Google Scholar

Lamers F, Vogelzangs N, Merikangas KR, de Jonge P, Beekman ATF, Penninx BWJH. 2013. Evidence for a differential role of HPA-axis function, inflammation and metabolic syndrome in melancholic versus atypical depression. Mol Psychiatry. 18(6):692–699.

 PubMed Web of Science ®Google Scholar

Michopoulos I, Zervas IM, Pantelis C, Tsaltas E, Papakosta V-M, Boufidou F, Nikolaou C, Papageorgiou C, Soldatos CR, Lykouras L. 2008. Neuropsychological and hypothalamic–pituitary-axis function in female patients with melancholic and non-melancholic depression. Eur Arch Psychiatry Clin Neurosci. 258(4):217–225.

 PubMed Web of Science ®Google Scholar

Veltman EM, Lamers F, Comijs HC, Stek ML, van der Mast RC, Rhebergen D. 2018. Inflammatory markers and cortisol parameters across depressive subtypes in an older cohort. J Affect Disord. 234:54–58.

 PubMed Web of Science ®Google Scholar

Maes M, Maes L, Schotte C, Vandewoude M, Martin M, D'Hondt P, Blockx P, Scharpe S, Cosyns P. 1990. Clinical subtypes of unipolar depression: part III. Quantitative differences in various biological markers between the cluster-analytically generated nonvital and vital depression classes. Psychiatry Res. 34(1):59–75.

 PubMed Web of Science ®Google Scholar

Hubain PP, Staner L, Dramaix M, Kerkhofs M, van Veeren C, Papadimitriou G, Mendlewicz J, Linkowski P. 1994. TSH response to TRH and EEG sleep in non-bipolar major depression: a multivariate approach. Eur Neuropsychopharmacol. 4(4):517–525.

 PubMed Web of Science ®Google Scholar

Gjerris A, Hammer M, Vendsborg P, Christensen NJ, Rafaelsen OJ. 1985. Cerebrospinal fluid vasopressin – changes in depression. Br J Psychiatry. 147:696–701.

 PubMed Web of Science ®Google Scholar

Maes M, Maes L, Schotte C, Cosyns P. 1992. A clinical and biological validation of the DSM-III melancholia diagnosis in men: results of pattern recognition methods. J Psychiatr Res. 26(3):183–196.

 PubMed Web of Science ®Google Scholar

Maes M, Meltzer HY, Cosyns P, Suy E, Schotte C. 1993. An evaluation of basal hypothalamic-pituitary-thyroid axis function in depression: results of a large-scaled and controlled study. Psychoneuroendocrinology. 18(8):607–620.

 PubMed Web of Science ®Google Scholar

Korgaonkar MS, Grieve SM, Koslow SH, Gabrieli JDE, Gordon E, Williams LM. 2011. Loss of white matter integrity in major depressive disorder: evidence using tract-based spatial statistical analysis of diffusion tensor imaging. Hum Brain Mapp. 32(12):2161–2171.

 PubMed Web of Science ®Google Scholar

Pizzagalli DA, Oakes TR, Fox AS, Chung MK, Larson CL, Abercrombie HC, Schaefer SM, Benca RM, Davidson RJ. 2004. Functional but not structural subgenual prefrontal cortex abnormalities in melancholia. Mol Psychiatry. 9(4):325–325.

 PubMed Web of Science ®Google Scholar

Exner C, Lange C, Irle E. 2009. Impaired implicit learning and reduced pre-supplementary motor cortex size in early-onset major depression with melancholic features. J Affect Disord. 119(1–3):156–162.

 PubMed Web of Science ®Google Scholar

Mertse N, Denier N, Walther S, Breit S, Grosskurth E, Federspiel A, Wiest R, Bracht T. 2022. Associations between anterior cingulate thickness, cingulum bundle microstructure, melancholia and depression severity in unipolar depression. J Affect Disord. 301:437–444.

 PubMed Web of Science ®Google Scholar

Takahashi T, Nishikawa Y, Yücel M, Whittle S, Lorenzetti V, Walterfang M, Sasabayashi D, Suzuki M, Pantelis C, Allen NB. 2016. Olfactory sulcus morphology in patients with current and past major depression. Psychiatry Res Neuroimaging. 255:60–65.

 PubMed Web of Science ®Google Scholar

Takahashi T, Sasabayashi D, Yücel M, Whittle S, Lorenzetti V, Walterfang M, Suzuki M, Pantelis C, Malhi GS, Allen NB. 2020. Pineal gland volume in major depressive and bipolar disorders. Front Psychiatry. 11:450.

 PubMed Web of Science ®Google Scholar

Hyett MP, Breakspear MJ, Friston KJ, Guo CC, Parker GB. 2015. Disrupted effective connectivity of cortical systems supporting attention and interoception in melancholia. JAMA Psychiatry. 72(4):350–358.

 PubMed Web of Science ®Google Scholar

Hyett MP, Parker GB, Guo CC, Zalesky A, Nguyen VT, Yuen T, Breakspear M. 2015. Scene unseen: disrupted neuronal adaptation in melancholia during emotional film viewing. Neuroimage Clin. 9:660–667.

 PubMed Web of Science ®Google Scholar

Hyett MP, Perry A, Breakspear M, Wen W, Parker GB. 2018. White matter alterations in the internal capsule and psychomotor impairment in melancholic depression. PLOS One. 13(4):e0195672.

 PubMed Web of Science ®Google Scholar

Guo CC, Hyett MP, Nguyen VT, Parker GB, Breakspear MJ. 2016. Distinct neurobiological signatures of brain connectivity in depression subtypes during natural viewing of emotionally salient films. Psychol Med. 46(7):1535–1545.

 PubMed Web of Science ®Google Scholar

Regonia PR, Takamura M, Nakano T, Ichikawa N, Fermin A, Okada G, Okamoto Y, Yamawaki S, Ikeda K, Yoshimoto J. 2021. Modelling heterogeneous brain dynamics of depression and melancholia using energy landscape analysis. Front Psychiatry. 12:780997.

 PubMed Web of Science ®Google Scholar

Workman CI, Lythe KE, McKie S, Moll J, Gethin JA, Deakin JFW, Elliott R, Zahn R. 2016. Subgenual cingulate–amygdala functional disconnection and vulnerability to melancholic depression. Neuropsychopharmacology. 41(8):2082–2090.

 PubMed Web of Science ®Google Scholar

Zhang Y, Cui X, Ou Y, Liu F, Li H, Chen J, Zhao J, Xie G, Guo W. 2021. Differentiating melancholic and non-melancholic major depressive disorder using fractional amplitude of low-frequency fluctuations. Front Psychiatry. 12:763770.

 PubMed Web of Science ®Google Scholar

Shan X, Cui X, Liu F, Li H, Huang R, Tang Y, Chen J, Zhao J, Guo W, Xie G. 2021. Shared and distinct homotopic connectivity changes in melancholic and non-melancholic depression. J Affect Disord. 287:268–275.

 PubMed Web of Science ®Google Scholar

Yan M, He Y, Cui X, Liu F, Li H, Huang R, Tang Y, Chen J, Zhao J, Xie G, et al. 2021. Disrupted regional homogeneity in melancholic and non-melancholic major depressive disorder at rest. Front Psychiatry. 12:618805.

 PubMed Web of Science ®Google Scholar

Yan M, Cui X, Liu F, Li H, Huang R, Tang Y, Chen J, Zhao J, Xie G, Guo W. 2021. Abnormal default-mode network homogeneity in melancholic and nonmelancholic major depressive disorder at rest. Neural Plast. 2021:6653309.

 PubMed Web of Science ®Google Scholar

Pizzagalli DA, Nitschke JB, Oakes TR, Hendrick AM, Horras KA, Larson CL, Abercrombie HC, Schaefer SM, Koger JV, Benca RM, et al. 2002. Brain electrical tomography in depression: the importance of symptom severity, anxiety, and melancholic features. Biol Psychiatry. 52(2):73–85.

 PubMed Web of Science ®Google Scholar

Nieber D, Schlegel S. 1992. Relationships between psychomotor retardation and EEG power spectrum in major depression. Neuropsychobiology. 25(1):20–23.

 PubMed Web of Science ®Google Scholar

Quinn CR, Rennie CJ, Harris AWF, Kemp AH. 2014. The impact of melancholia versus non-melancholia on resting-state, EEG alpha asymmetry: eelectrophysiological evidence for depression heterogeneity. Psychiatry Res. 215(3):614–617.

 PubMed Web of Science ®Google Scholar

Tsujii N, Mikawa W, Akashi H, Tsujimoto E, Adachi T, Kirime E, Takaya M, Yanagi M, Shirakawa O. 2014. Right temporal activation differs between melancholia and nonmelancholic depression: a multichannel near-infrared spectroscopy study. J Psychiatr Res. 55:1–7.

 PubMed Web of Science ®Google Scholar

Tsujii N, Mikawa W, Tsujimoto E, Akashi H, Adachi T, Kirime E, Shirakawa O. 2016. Relationship between prefrontal hemodynamic responses and quality of life differs between melancholia and non-melancholic depression. Psychiatry Res Neuroimaging. 253:26–35.

 PubMed Web of Science ®Google Scholar

Foti D, Carlson JM, Sauder CL, Proudfit GH. 2014. Reward dysfunction in major depression: multimodal neuroimaging evidence for refining the melancholic phenotype. Neuroimage. 101:50–58.

 PubMed Web of Science ®Google Scholar

Shankman SA, Sarapas C, Klein DN. 2011. The effect of pre- vs. post-reward attainment on EEG asymmetry in melancholic depression. Int J Psychophysiol. 79(2):287–295.

 PubMed Web of Science ®Google Scholar

Liu H, Sarapas C, Shankman SA. 2016. Anticipatory reward deficits in melancholia. J Abnorm Psychol. 125(5):631–640.

 PubMed Web of Science ®Google Scholar

Biver F, Goldman S, Delvenne V, Luxen A, De Maertelaer V, Hubain P, Mendlewicz J, Lotstra F. 1994. Frontal and parietal metabolic disturbances in unipolar depression. Biol Psychiatry. 36(6):381–388.

 PubMed Web of Science ®Google Scholar

Austin MP, Dougall N, Ross M, Murray C, O'Carroll RE, Moffoot A, Ebmeier KP, Goodwin GM. 1992. Single photon emission tomography with 99mTc-exametazime in major depression and the pattern of brain activity underlying the psychotic/neurotic continuum. J Affect Disord. 26(1):31–43.

 PubMed Web of Science ®Google Scholar

Fountoulakis KN, Iacovides A, Gerasimou G, Fotiou F, Ioannidou C, Bascialla F, Grammaticos P, Kaprinis G. 2004. The relationship of regional cerebral blood flow with subtypes of major depression. Prog Neuropsychopharmacol Biol Psychiatry. 28(3):537–546.

 PubMed Web of Science ®Google Scholar

Elton M. 1984. A longitudinal investigation of event-related potentials in depression. Biol Psychiatry. 19(12):1635–1649.

 PubMed Web of Science ®Google Scholar

Quinn CR, Harris A, Kemp AH. 2012. The impact of depression heterogeneity on inhibitory control. Aust N Z J Psychiatry. 46(4):374–383.

 PubMed Web of Science ®Google Scholar

Weinberg A, Liu H, Shankman SA. 2016. Blunted neural response to errors as a trait marker of melancholic depression. Biol Psychol. 113:100–107.

 PubMed Web of Science ®Google Scholar

Fitzgerald PB, Mellow TB, Hoy KE, Segrave R, Cooper NR, Upton DJ, Croft RJ. 2009. A study of intensity dependence of the auditory evoked potential (IDAEP) in medicated melancholic and non-melancholic depression. J Affect Disord. 117(3):212–216.

 PubMed Web of Science ®Google Scholar

Kerr CC, Kemp AH, Rennie CJ, Robinson PA. 2011. Thalamocortical changes in major depression probed by deconvolution and physiology-based modeling. Neuroimage. 54(4):2672–2682.

 PubMed Web of Science ®Google Scholar

Schlegel S, Nieber D, Herrmann C, Bakauski E. 1991. Latencies of the P300 component of the auditory event-related potential in depression are related to the Bech-Rafaelsen Melancholia Scale but not to the Hamilton Rating Scale for depression. Acta Psychiatr Scand. 83(6):438–440.

 PubMed Web of Science ®Google Scholar

Kemp AH, Pe Benito L, Quintana DS, Clark CR, McFarlane A, Mayur P, Harris A, Boyce P, Williams LM. 2010. Impact of depression heterogeneity on attention: an auditory oddball event related potential study. J Affect Disord. 123(1-3):202–207.

 PubMed Web of Science ®Google Scholar

Fotiou F, Fountoulakis KN, Iacovides A, Kaprinis G. 2003. Pattern-reversed visual evoked potentials in subtypes of major depression. Psychiatry Res. 118(3):259–271.

 PubMed Web of Science ®Google Scholar

Winograd-Gurvich C, Georgiou-Karistianis N, Fitzgerald PB, Millist L, White OB. 2006a. Ocular motor differences between melancholic and non-melancholic depression. J Affect Disord. 93(1–3):193–203.

 PubMed Web of Science ®Google Scholar

Winograd-Gurvich C, Georgiou-Karistianis N, Fitzgerald PB, Millist L, White OB. 2006b. Self-paced and reprogrammed saccades: differences between melancholic and non-melancholic depression. Neurosci Res. 56(3):253–260.

 PubMed Web of Science ®Google Scholar

Fountoulakis KN, Fotiou F, Iacovides A, Kaprinis G. 2005. Is there a dysfunction in the visual system of depressed patients? Ann Gen Psychiatry. 4(1):7.

 PubMedGoogle Scholar

Giles DE, Roffwarg HP, Schlesser MA, Rush AJ. 1986. Which endogenous depressive symptoms relate to REM latency reduction? Biol Psychiatry. 21(5–6):473–482.

 PubMed Web of Science ®Google Scholar

Giles DE, Roffwarg HP, Rush AJ, Guzick DS. 1990. Age-adjusted threshold values for reduced REM latency in unipolar depression using ROC analysis. Biol Psychiatry. 27(8):841–853.

 PubMed Web of Science ®Google Scholar

Hubain PP, Souery D, Jönck L, Staner L, Van Veeren C, Kerkhofs M, Mendlewicz J, Linkowski P. 1995. Relationship between the newcastle scale and sleep polysomnographic variables in major depression: a controlled study. Eur Neuropsychopharmacol. 5(2):129–134.

 PubMed Web of Science ®Google Scholar

Riemann D, Hohagen F, Bahro M, Berger M. 1994. Sleep in depression: the influence of age, gender and diagnostic subtype on baseline sleep and the cholinergic REM induction test with RS 86. Eur Arch Psychiatry Clin Neurosci. 243(5):279–290.

 PubMed Web of Science ®Google Scholar

Hein M, Lanquart J-P, Loas G, Hubain P, Linkowski P. 2019. Alterations of neural network organisation during rapid eye movement sleep and slow-wave sleep in major depression: implications for diagnosis, classification, and treatment. Psychiatry Res Neuroimag. 291:71–78.

 PubMed Web of Science ®Google Scholar

Patas K, Penninx BWJH, Bus BAA, Vogelzangs N, Molendijk ML, Elzinga BM, Bosker FJ, Oude Voshaar RC. 2014. Association between serum brain-derived neurotrophic factor and plasma interleukin-6 in major depressive disorder with melancholic features. Brain Behav Immun. 36:71–79.

 PubMed Web of Science ®Google Scholar

Spanemberg L, Caldieraro MA, Vares EA, Wollenhaupt-Aguiar B, Kauer-Sant’Anna M, Kawamoto SY, Galvao E, Parker G, Fleck MP. 2014. Biological differences between melancholic and nonmelancholic depression subtyped by the CORE measure. Neuropsychiatr Dis Treat. 10:1523–1531.

 PubMed Web of Science ®Google Scholar

Lehto S, Tolmunen T, Joensuu M, Saarinen PI, Vanninen R, Ahola P, Tiihonen J, Kuikka J, Lehtonen J. 2006. Midbrain binding of [123I]nor-β-CIT in atypical depression. Prog Neuropsychopharmacol Biol Psychiatry. 30(7):1251–1255.

 PubMed Web of Science ®Google Scholar

Glaus J, Vandeleur CL, von Känel R, Lasserre AM, Strippoli M-PF, Gholam-Rezaee M, Castelao E, Marques-Vidal P, Bovet P, Merikangas K, et al. 2014. Associations between mood, anxiety or substance use disorders and inflammatory markers after adjustment for multiple covariates in a population-based study. J Psychiatr Res. 58:36–45.

 PubMed Web of Science ®Google Scholar

Lamers F, Bot M, Jansen R, Chan MK, Cooper JD, Bahn S, Penninx BWJH. 2016. Serum proteomic profiles of depressive subtypes. Transl Psychiatry. 6(7):e851.

 PubMed Web of Science ®Google Scholar

Lasserre AM, Strippoli M-PF, Glaus J, Gholam-Rezaee M, Vandeleur CL, Castelao E, Marques-Vidal P, Waeber G, Vollenweider P, Preisig M. 2017. Prospective associations of depression subtypes with cardio-metabolic risk factors in the general population. Mol Psychiatry. 22(7):1026–1034.

 PubMed Web of Science ®Google Scholar

Liu H, Wu X, Wang Y, Liu X, Peng D, Wu Y, Chen J, Su Y, Xu J, Ma X, et al. 2022. TNF-α, IL-6 and hsCRP in patients with melancholic, atypical and anxious depression: an antibody array analysis related to somatic symptoms. Gen Psychiatr. 35(4):e100844.

 PubMed Web of Science ®Google Scholar

Mohamed AE, El-Latif RRA, Youssef AM, Ibrahim AS. 2020. C-reactive protein and clinical subtypes of major depressive disorder at Zagazig University Hospitals. Middle East Curr Psychiatry. 27(1):35.

 Google Scholar

Rothermundt M, Arolt V, Peters M, Gutbrodt H, Fenker J, Kersting A, Kirchner H. 2001. Inflammatory markers in major depression and melancholia. J Affect Disord. 63(1–3):93–102.

 PubMed Web of Science ®Google Scholar

Rudaz DA, Vandeleur CL, Gebreab SZ, Gholam-Rezaee M, Strippoli M-PF, Lasserre AM, Glaus J, Castelao E, Pistis G, von Känel R, et al. 2017. Partially distinct combinations of psychological, metabolic and inflammatory risk factors are prospectively associated with the onset of the subtypes of major depressive disorder in midlife. J Affect Disord. 222:195–203.

 PubMed Web of Science ®Google Scholar

Sánchez-Carro Y, de la Torre-Luque A, Leal-Leturia I, Salvat-Pujol N, Massaneda C, de Arriba-Arnau A, Urretavizcaya M, Pérez-Solà V, Toll A, Martínez-Ruiz A, et al. 2023. Importance of immunometabolic markers for the classification of patients with major depressive disorder using machine learning. Prog Neuropsychopharmacol Biol Psychiatry. 121:110674.

 PubMed Web of Science ®Google Scholar

Vogelzangs N, Comijs HC, Oude Voshaar RC, Stek ML, Penninx BWJH. 2014. Late-life depression symptom profiles are differentially associated with immunometabolic functioning. Brain Behav Immun. 41:109–115.

 PubMed Web of Science ®Google Scholar

Dunjic-Kostic B, Ivkovic M, Radonjic NV, Petronijevic ND, Pantovic M, Damjanovic A, Poznanovic ST, Jovanovic A, Nikolic T, Jasovic-Gasic M. 2013. Melancholic and atypical major depression—connection between cytokines, psychopathology and treatment. Prog Neuropsychopharmacol Biol Psychiatry. 43:1–6.

 PubMed Web of Science ®Google Scholar

Huang T-L, Lee C-T. 2007. T-helper 1/T-helper 2 cytokine imbalance and clinical phenotypes of acute-phase major depression. Psychiatry Clin Neurosci. 61(4):415–420.

 PubMed Web of Science ®Google Scholar

Maes M, Mihaylova I, Kubera M, Ringel K. 2012. Activation of cell-mediated immunity in depression: association with inflammation, melancholia, clinical staging and the fatigue and somatic symptom cluster of depression. Prog Neuropsychopharmacol Biol Psychiatry. 36(1):169–175.

 PubMed Web of Science ®Google Scholar

Marques-Deak AH, Neto FL, Dominguez WV, Solis AC, Kurcgant D, Sato F, Ross JM, Prado EBA. 2007. Cytokine profiles in women with different subtypes of major depressive disorder. J Psychiatr Res. 41(1–2):152–159.

 PubMed Web of Science ®Google Scholar

Kaestner F, Hettich M, Peters M, Sibrowski W, Hetzel G, Ponath G, Arolt V, Cassens U, Rothermundt M. 2005. Different activation patterns of proinflammatory cytokines in melancholic and non-melancholic major depression are associated with HPA axis activity. J Affect Disord. 87(2–3):305–311.

 PubMed Web of Science ®Google Scholar

Sowa-Kućma M, Styczeń K, Siwek M, Misztak P, Nowak RJ, Dudek D, Rybakowski JK, Nowak G, Maes M. 2018. Lipid peroxidation and immune biomarkers are associated with major depression and its phenotypes, including treatment-resistant depression and melancholia. Neurotox Res. 33(2):448–460.

 PubMed Web of Science ®Google Scholar

Rothermundt M, Arolt V, Fenker J, Gutbrodt H, Peters M, Kirchner H. 2001. Different immune patterns in melancholic and non-melancholic major depression. Eur Arch Psychiatry Clin Neurosci. 251(2):90–97.

 PubMed Web of Science ®Google Scholar

Erdem M, Celik C, Cayci T, Ozdemir B, Kurt YG, Akgul EO, Yaman H, Balikci A, Uzun O. 2011. Serum haptoglobin levels in patients with melancholic and nonmelancholic major depression. Prog Neuropsychopharmacol Biol Psychiatry. 35(4):944–947.

 PubMed Web of Science ®Google Scholar

Maes M, Scharpé S, Meltzer HY, Okayli G, Bosmans E, D'Hondt P, Vanden Bossche BV, Cosyns P. 1994. Increased neopterin and interferon-gamma secretion and lower availability of L-tryptophan in major depression: further evidence for an immune response. Psychiatry Res. 54(2):143–160.

 PubMed Web of Science ®Google Scholar

Maes M, Ringel K, Kubera M, Berk M, Rybakowski J. 2012. Increased autoimmune activity against 5-HT: a key component of depression that is associated with inflammation and activation of cell-mediated immunity, and with severity and staging of depression. J Affect Disord. 136(3):386–392.

 PubMed Web of Science ®Google Scholar

Eriksson MD, Eriksson JG, Korhonen P, Koponen H, Salonen MK, Mikkola TM, Kajantie E, Wasenius NS, von Bonsdorff M, Kautiainen H, et al. 2023. Depressive symptoms and mortality-findings from helsinki birth cohort study. Acta Psychiatr Scand. 147(2):175–185.

 PubMed Web of Science ®Google Scholar

Huang T-L. 2005. Serum lipid profiles in major depression with clinical subtypes, suicide attempts and episodes. J Affect Disord. 86(1):75–79.

 PubMed Web of Science ®Google Scholar

Selenius JS, Silveira PP, Salonen M, Kautiainen H, von Bonsdorff M, Kajantie E, Lahti J, Eriksson JG, Wasenius NS. 2021. The relationship between health-related quality of life and melancholic depressive symptoms is modified by brain insulin receptor gene network. Sci Rep. 11(1):21588.

 PubMed Web of Science ®Google Scholar

Maes M, Scharpé S, Meltzer HY, Suy E, Cosyns P, Calabrese J. 1992. Lower angiotensin I converting enzyme activity in melancholic subjects: a pilot study. Biol Psychiatry. 32(7):621–624.

 PubMed Web of Science ®Google Scholar

Maes M, D'Haese PC, Scharpé S, D'Hondt P, Cosyns P, De Broe ME. 1994. Hypozincemia in depression. J Affect Disord. 31(2):135–140.

 PubMed Web of Science ®Google Scholar

Styczeń K, Sowa-Kućma M, Siwek M, Dudek D, Reczyński W, Szewczyk B, Misztak P, Topór-Mądry R, Opoka W, Nowak G. 2017. The serum zinc concentration as a potential biological marker in patients with major depressive disorder. Metab Brain Dis. 32(1):97–103.

 PubMed Web of Science ®Google Scholar