Abstract

Major depressive disorder is a clinical syndrome diagnosed when a minimum number of symptoms persist over a minimum period, with low mood or a loss of interest or pleasure among them. MeSH classifies major depressive disorder under depressive disorder, within mood disorders. Patients satisfy its polythetic criteria through many symptom combinations, and cognitive accounts tie it to negative biases in attention, interpretation and memory, weaker cognitive control and poorer reward processing. Whether those biases cause the disorder or follow from it remains open, and biological findings, from serotonin to inflammation and brain imaging, are small, noisy or contested at the group level. Psychotherapy, antidepressants, their combination and brain stimulation all help, although publication bias inflated published effects and whether benefit depends on severity is disputed; maintenance medication and mindfulness-based cognitive therapy are the main tested routes to preventing relapse.

Keywords: major depressive disorder, depression specifiers, cognitive model of depression, cognitive bias, antidepressant treatment

Major depressive disorder is diagnosed when at least five symptoms are present together over the same two weeks, mark a change from the person's earlier functioning and include low mood or a loss of interest or pleasure (MeSH, D003865). This page covers the diagnosed disorder; depressive mood and depression as a symptom across conditions are covered on the depression page. The disorder is both common and severely disabling, and it varies widely in presentation, course and response to treatment (Malhi & Mann, 2018). In a national household survey of 9090 US adults, lifetime prevalence reached 16.2% and 12-month prevalence 6.6%, yet among people with the disorder in the past 12 months who were treated in health care, only 41.9% received care that met the survey's adequacy criteria (Kessler et al., 2003). The disorder matters to cognitive psychology because its leading psychological accounts place much of it in how information about the self is attended to, interpreted and remembered, and because its cognitive costs can be measured in their own right.

The article first defines the disorder in What Major Depressive Disorder Is, traces the History of the Concept, and describes how criteria and rating scales are applied in Diagnosing and Measuring Major Depressive Disorder. The Experience of Major Depressive Disorder draws on qualitative accounts of living with depression, of deciding to seek help and of taking antidepressants, and Types of Major Depressive Disorder takes each course descriptor, specifier and boundary presentation in turn, with its cognitive findings where any exist. Prevalence, Course and Burden sets out the epidemiology. The psychological theories follow in Beck's Cognitive Model of Depression, which also asks whether cognitive change carries part of the effect of cognitive therapy, Helplessness, Hopelessness and Rumination and Interpersonal and Evolutionary Accounts, and the laboratory evidence in Attention, Memory and Executive Biases and Reward Processing and Anhedonia. The second of these separates awaiting a reward from receiving it and asks whether anhedonia concerns motivation or pleasure. Cause or Consequence? asks whether these features precede the disorder, and Validity of Laboratory Measures asks how well the tasks measure, what experience sampling adds and how far test deficits bear on everyday functioning. Stress, Adversity and Onset, which covers life events, childhood maltreatment and the test of a gene-by-stress interaction, The Biology of Major Depressive Disorder, Sleep and Circadian Rhythms, which treats insomnia as a predictor of onset and reviews sleep-based treatments, and Brain Networks and Imaging cover environmental and biological accounts, and Heterogeneity and Symptom Networks questions whether a single diagnosis is the right unit. Culture and Social Context, Sex Differences, Major Depressive Disorder Across the Lifespan, Comorbidity, Depression in the Family, Cognitive Impairment and Dementia in Later Life, Suicide Risk and Related Conditions and Normal Sadness place the disorder among people and neighbouring conditions; the culture section covers symptom presentation, idioms of distress, socioeconomic inequality and barriers to seeking help, and the suicide section covers the limits of risk prediction, defeat and entrapment, and safety planning. Within that group, the sex differences section measures the female excess in diagnoses and weighs an explanatory review, the lifespan section follows the disorder from youth to old age, the comorbidity section covers co-occurring physical and mental disorders and the genetic liability depression shares with generalized anxiety, the family section covers the children of depressed mothers and relatives' own accounts, and the later-life section asks whether depressive symptoms are a risk factor for dementia or an early sign of it and how cognitive deficits relate to functioning. The final sections review Treatment Evidence, including the dispute over severity and the reanalysis of the STAR*D trial, How Psychological Treatments Work, Brain Stimulation and Rapid-Acting Treatments and Preventing Relapse and Recurrence, which weighs maintenance medication against mindfulness-based cognitive therapy; a Worked Example counts the symptom combinations the criteria allow, and Current Directions closes the body.

Key Takeaways

  • The diagnosis is polythetic: in interviews with more than 1,500 patients diagnosed under DSM-IV criteria, 170 of the 227 symptom combinations the criteria allow actually occurred, and only a few were frequent enough to serve as prototypes.
  • Specifiers describe recognised presentations rather than proven separate diseases, and their cognitive profiles differ: psychotic depression showed poorer overall cognition than non-psychotic depression across 12 studies, and melancholic depression somewhat poorer cognition across 9 studies, a difference not detected among drug-free patients. Cognition did not cleanly separate unipolar from bipolar patients, who did not differ during depressive episodes.
  • Meta-analyses link depression to a moderate negative attentional bias that was clear on the dot probe but only marginal on the emotional Stroop, and to a medium-sized interpretation bias that appeared on direct but not indirect measures; one review concluded that weak cognitive control probably underlies negative biases in attention, interpretation and memory.
  • In a meta-analysis of CANTAB studies, executive and attentional deficits persisted in remitted patients while the memory difference was non-significant; a small single-site study suggested that momentary influences such as well-being make deficits look larger than they are. A meta-analysis of 113 studies found executive impairment across measures, and in one cross-sectional study currently depressed adults inhibited negative material less well than formerly and never-depressed adults. Reviews linked test deficits to poorer psychosocial functioning, although one of them rated that evidence weak. Late-life depression was associated with a higher risk of later dementia across 23 cohort studies, an association rather than a demonstrated cause; in the 28-year Whitehall II cohort, symptoms recorded in later life predicted dementia while symptoms recorded in midlife did not, so late-life symptoms may be an early sign of dementia, or share causes with it, rather than a risk factor in their own right.
  • Reward processing was poorer in depressed groups by a small-to-medium margin across 48 case-control studies; in one imaging study the blunted striatal response appeared mainly when rewards arrived, with weaker evidence while they were awaited, and one review argued that dopamine bears more on motivation than on pleasure, so the definition of anhedonia needs refining.
  • Beck's cognitive model is supported as a description of depressive thinking and only partly as a causal account; in cognitive therapy, cognitive change tended to precede later symptom change, though few studies ran formal mediation tests. The causal evidence is mixed: negative cognitive styles preceded onset in a high-risk student cohort, a Dutch population cohort found vulnerability raised before a first episode with no scar effect afterwards, and cognitive bias modification produced small, unreliable benefits with insufficient evidence for depression, so whether depressive cognition is cause or consequence remains open.
  • Biological differences such as raised inflammatory markers, HPA axis hyperactivity and lower serum BDNF in untreated patients appear at the group level, but brain differences were tiny and overlapping in a large imaging cohort, the serotonin hypothesis is disputed, and historical candidate genes found no support in very large samples. Pooled across 361 studies, depressed people had higher cortisol and ACTH but no significant CRH difference. On the environmental side, stressful life events predicted onset in a co-twin analysis of women, and childhood maltreatment was linked to recurrent or persistent depression and to poorer treatment outcome, while a collaborative analysis of 38,802 European-ancestry subjects found no strong interaction between 5-HTTLPR genotype and stress.
  • Insomnia predicted later depression across 21 longitudinal studies; in patients with another mental disorder, cognitive behavioural therapy for insomnia eased insomnia and depressive symptoms after treatment, though effects on comorbid symptoms were not significant at follow-up, and adding sleep deprivation to other treatments gave no significant extra benefit.
  • Culture and social position shape presentation and care: Chinese outpatients reported more somatic symptoms on two of three methods and Euro-Canadian outpatients more psychological symptoms on all three, the lowest socioeconomic groups had higher odds of depression, and stigma had a small-to-moderate negative effect on help-seeking. Across 24 countries, low perceived need and attitudinal barriers were the dominant reasons for not seeking or for stopping treatment, and in qualitative syntheses asking for help was experienced as a threat to identity, while people taking antidepressants often felt the drugs were the only option open to them.
  • Women had nearly twice the odds of a diagnosis, a gap larger in adolescence than in adulthood, although one review judged the evidence for depression-specific risk factors limited. Depression was most strongly associated with generalized anxiety disorder in a US national survey, with which it shared a genetic correlation of +1.00 in women and +0.74 in men in one twin study; children of depressed mothers showed more emotional and behavioural problems, though every association was small; and subthreshold depression roughly doubled the incidence of later major depression in youth, adults and older people alike.
  • Broad risk factors predicted suicidal thoughts and behaviours only slightly better than chance across 50 years of studies; the integrated motivational-volitional model, in which defeat leads to ideation through entrapment, has growing but unresolved support, and safety planning-type interventions reduced suicidal behaviour, though not significantly ideation, in a small evidence base of low methodological quality.
  • All 21 antidepressants in one network meta-analysis outperformed placebo, psychotherapy and medication did not differ from each other, and combined treatment produced more responders than either alone, while comparisons with FDA records and unpublished funded trials showed that published reports overstated both drug and psychotherapy effects. Analyses disagree on whether drug benefit depends on baseline severity, a reanalysis of STAR*D put cumulative remission at 35.0% rather than the reported 67%, ECT's cognitive deficits were concentrated in the first 3 days after treatment, and ketamine was noninferior to ECT, with less memory decline, in one open-label trial. To prevent relapse, stopping maintenance antidepressants led to more relapse than continuing them in a primary care trial, and mindfulness-based cognitive therapy lowered relapse risk in pooled patient data but was not superior to maintenance medication in a head-to-head trial.

What Major Depressive Disorder Is

MeSH defines major depressive disorder by a count and a time window: at least five symptoms present together over the same two weeks, marking a change from the person's earlier functioning, with low mood or a loss of interest or pleasure among them; its scope note credits this definition to DSM-5 (MeSH, D003865). MeSH lists clinical depression and major depression disorder among its entry terms, together with four historical terms: involutional depression, involutional melancholia, involutional paraphrenia and involutional psychosis.

The diagnosis names a clinical syndrome, not the passing low mood or sadness that follows ordinary loss and disappointment; where the boundary falls, and how it is drawn in practice, is taken up under Related Conditions and Normal Sadness. Current clinical criteria, and the specifiers that capture an episode's features and course, come from DSM-5-TR, a text revision of the fifth Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2022). The section Diagnosing and Measuring Major Depressive Disorder describes how those criteria are applied.

Major depression is both common and severely disabling, and it erodes psychosocial functioning and quality of life (Malhi & Mann, 2018). That Lancet seminar also stresses how unevenly the disorder behaves from one person to the next: presentations vary, the course and prognosis are hard to predict, and response to treatment differs widely (Malhi & Mann, 2018). A disease primer framed the disorder as multifactorial (Otte et al., 2016). Its authors traced risk to heritable influences and to childhood abuse, noted that women receive the diagnosis about twice as often as men, and reviewed candidate pathways running through the hypothalamic-pituitary-adrenal axis, hippocampal volume, and the cognitive control and affective-salience networks; they nonetheless judged that no established mechanism can account for every aspect of the disorder (Otte et al., 2016). These strands are developed in The Biology of Major Depressive Disorder, Brain Networks and Imaging, and Sex Differences.

A further question is whether the disorder is a category at all. A definition built on a symptom threshold treats people as either having the disorder or not, but taxometric methods can test whether a latent variable is truly categorical (a taxon) or varies by degree along a dimension. A quantitative review of this literature pooled 311 distinct findings from 177 articles, covering 533 377 participants, and concluded that most latent variables in psychopathology are dimensional (Haslam et al., 2012). Taxonic results were less frequent in recent studies than in early ones, less frequent where methods were stronger, and less frequent where comparison data fit indices were used, which led the authors to judge that many early taxa were likely spurious (Haslam et al., 2012). Mood disorders were among the domains covered and sided with the dimensional majority; the authors left the evidence open, as not definitive, only for schizotypy, autism and substance use disorders (Haslam et al., 2012). The diagnostic cut-off is therefore best read as a threshold on a continuum rather than a boundary found in nature, a point that returns in Heterogeneity and Symptom Networks.

History of the Concept

MeSH's own history note records the shift in vocabulary: the heading Major Depressive Disorder dates from 1981, and from 1963 to 1980 the literature was indexed under involutional psychoses (MeSH, D003865).

That indexing change mirrors a change in clinical thinking. For much of the twentieth century American psychiatry recognised involutional melancholia, a diagnosis applied chiefly to older patients, most of them women past menopause. A historical analysis of the American psychiatric literature from 1900 to 1980 traced how this age- and sex-linked category lost its separate identity and was absorbed into major depression (Hirshbein, 2009). The analysis is historical rather than epidemiological, so it documents how clinicians used the label, not how common the condition was.

The same decades brought the first attempts to measure depression rather than only describe it. Hamilton's 1960 paper set out a rating scale for depression (Hamilton, 1960). A year later, Beck and colleagues published an inventory for measuring depression, against a background of low agreement between clinicians on psychiatric diagnosis (Beck et al., 1961). Beck then gave depression a book-length treatment covering its clinical, experimental and theoretical aspects (Beck, 1967); the cognitive account that grew from this work is discussed under Beck's Cognitive Model of Depression, and the current use of rating instruments under Diagnosing and Measuring Major Depressive Disorder.

Operational criteria trace back to a proposal setting out explicit criteria for 14 psychiatric illnesses, which its authors offered as the best then available and expected to be tested and revised (Feighner et al., 1972). The Research Diagnostic Criteria came next, and their first reliability studies reported that diagnoses made with them were highly reliable (Spitzer et al., 1978). Critics have asked what standardised criteria left behind. Kendler compared the symptoms and signs of depression described in psychiatric textbooks published between 1900 and 1960 with the symptomatic criteria of DSM-5 (Kendler, 2016). The older textbooks gave more weight to cognitive, physical and psychomotor features, whereas the DSM criteria leaned more heavily on neurovegetative symptoms and covered a narrower range of mood states and cognitive content. Kendler concluded that important features of the classic descriptions, including subtle phenomena that are hard to assess, are not captured by the modern list, and that the criteria have been reified: in his view the DSM should index psychiatric disorders rather than constitute them.

A cross-cultural episode showed that the boundaries of the concept also depended on where it was applied. Kleinman's field study combined clinical and anthropological methods and followed 100 outpatients in Hunan, China, who carried a diagnosis of neurasthenia; when he re-examined them against DSM-III criteria, most turned out to have depression that showed itself as bodily complaints (Kleinman, 1982). Symptoms improved more than either help-seeking declined or perceived social impairment eased, although improvement was judged from the researcher's side rather than the patients'. Kleinman proposed that neurasthenia is best understood as a biocultural illness experience, a theme taken up in Culture and Social Context.

Diagnosing and Measuring Major Depressive Disorder

A diagnosis requires meeting the DSM-5-TR criteria, DSM-5-TR being the text-revised fifth Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2022). In outline, the manual requires that a minimum number of symptoms persist over a minimum period, that at least one of them be one of two core symptoms, and that the episode bring clinically significant distress or impair functioning (American Psychiatric Association, 2022). The criteria are paraphrased here rather than reproduced; the MeSH definition summarised under What Major Depressive Disorder Is gives the symptom count and the two-week window (MeSH, D003865).

Because the definition is polythetic, two people can both meet it while sharing few symptoms. A polythetic category is satisfied by any sufficient subset of its criteria rather than by one required set, and the practical question is how much of that latitude real patients use. Zimmerman and colleagues answered it with semi-structured interviews of more than 1,500 patients diagnosed with DSM-IV major depressive disorder: of the 227 symptom combinations the criteria allow, 170 occurred in the sample, and the remaining 57 did not (Zimmerman et al., 2015). They concluded that the diagnosis is highly heterogeneous and that only a few combinations recur often enough to count as prototypes (Zimmerman et al., 2015).

One diagnosis, many profiles

Pick symptoms for one person. The count rule needs at least five, one of them a core symptom (heavier border).

Symptoms selected5 of 9
Core symptom presentyes
Meets the count ruleyes
Shared with a second profile (interest, appetite, sleep, movement, fatigue)1 of 5
Combinations that meet the rule227

Both profiles qualify, yet they share at most one symptom.

Arithmetic illustration, not an assessment, using a simplified count rule (five of nine symptoms, at least one of two core symptoms) with symptom labels paraphrased. The rule allows 227 combinations; in a large clinical sample, 170 of the 227 were observed (Zimmerman et al., 2015). The real criteria also set a duration and require distress or impairment.

Severity is measured with instruments of two kinds, those the patient completes and those a clinician scores. Beck and colleagues introduced the Beck Depression Inventory in an instrument-development paper that pointed to the low agreement clinicians reached on psychiatric diagnosis (Beck et al., 1961). The Hamilton Rating Scale for Depression is instead a clinician-rated scale, published by Hamilton as a rating scale for depression (Hamilton, 1960). The two are not interchangeable in content, since the Hamilton scale gives somatic items more weight than the Beck inventory does (Fried & Nesse, 2015).

The PHQ-9 is a brief self-report severity measure from the Patient Health Questionnaire, its items built on the 9 DSM-IV criteria. Kroenke and colleagues validated it in 6,000 patients from 8 primary care clinics and 7 obstetrics-gynecology clinics, 580 of whom also had a structured interview with a mental health professional (Kroenke et al., 2001). Against that interview, a score of 10 or more detected major depression with a sensitivity of 88% and a specificity of 88% (Kroenke et al., 2001). Higher scores went with poorer function on the 20-item Short-Form General Health Survey, more sick days, more health care use and greater difficulty from symptoms, and the two clinic types gave similar results (Kroenke et al., 2001). The authors judged the PHQ-9 a valid measure of severity and marked thresholds at scores of 5, 10, 15 and 20 (Kroenke et al., 2001). Both validation samples were medical clinic patients, not psychiatric ones.

Each of these instruments is usually reduced to a single total, and that habit has drawn a sustained critique. In a narrative review, Fried and Nesse argued that depressive symptoms are distinct from one another and that adding them into a sum score hides information (Fried & Nesse, 2015). The review also noted that the field has produced 280 different depression scales, and it recommended analysing symptoms individually, testing causal links among them and moving towards treatment matched to the person (Fried & Nesse, 2015). The same authors conceded that existing scales were not designed for symptom-level analysis, that sum scores still carry general information and are less exposed to some measurement problems, and that some of the results they drew on require replication (Fried & Nesse, 2015). The argument and its evidence are taken up in Heterogeneity and Symptom Networks.

The Experience of Major Depressive Disorder

Diagnostic criteria list symptoms; qualitative research asks what those symptoms are like to live with. The evidence in this section comes from interview and synthesis studies, not from measurement, and it describes depression as its participants reported it rather than testing any account of its cause.

The broadest synthesis concerns young people. A qualitative systematic review pooled 39 studies with 884 adolescents and young people aged 10 to 24 years, drawn from 16 countries, and organised their accounts through thematic synthesis into 3 themes (Viduani et al., 2024). Across those accounts, the features participants mentioned most often were sadness, pulling back from other people and loneliness (Viduani et al., 2024). The reviewers concluded that features adolescents themselves describe could potentially make detection more accurate, a suggestion about detection rather than a demonstrated gain (Viduani et al., 2024). Two limits bound how far the picture travels: most of the included studies came from high-income countries, and most participants were female (Viduani et al., 2024). The review concerns depression as young people reported it, not a confirmed diagnosis of major depressive disorder in every participant, and it says nothing about adults.

A much smaller interview study looked closely at a first episode in adults. Using semi-structured interviews analysed with interpretative phenomenological analysis, Smith and Rhodes spoke with seven people in London who were in their first episode of depression (Smith & Rhodes, 2015). Their participants described an unstable pattern in which a major depletion was interrupted by bouts of emotional and cognitive agitation (Smith & Rhodes, 2015). The depletion reached three domains of experience, so that relationships, the body and the felt passage of time were each diminished, and the participant's whole life-world seemed to shrink (Smith & Rhodes, 2015). Participants also reported a loss event before the episode began, an association they described themselves rather than one the study tested (Smith & Rhodes, 2015). The authors read the first episode as an existential threat to the self and recommended that therapy explore that meaning early (Smith & Rhodes, 2015). With seven participants from one city, the study offers depth of description, not prevalence, and its themes are the analysts' interpretation of what participants said.

Taken together, the two studies sketch depression from the inside as a narrowing of social contact, bodily vitality and time, felt as a threat to the self, alongside low mood and isolation. Neither design can say how common any of these experiences is among people meeting the criteria set out in Diagnosing and Measuring Major Depressive Disorder.

Seeking Help and Taking Antidepressants

Qualitative syntheses have also followed people past the onset of the episode, to the decision to seek help and to life on medication. A meta-ethnography of 20 papers, covering 474 people with depression, identified help-seeking as a threat to identity as one of its major concepts (Doblyte & Jiménez-Mejías, 2017). Relying on alternative coping strategies emerged as the main factor in delaying treatment, while the person's social network could either encourage help-seeking or bring conflict over it (Doblyte & Jiménez-Mejías, 2017).

For people who do start treatment, a thematic meta-synthesis of 27 studies with 2937 participants examined what taking antidepressants is like (Crowe et al., 2023). Most participants felt the medication was the only option open to them, and participants described the explanation of depression as a biochemical deficit as stigmatising (Crowe et al., 2023). Accounts diverged: some participants found the drugs helpful, whereas others described their feelings as masked and their sense of self altered, a sense of being not myself, and coming off the medication formed part of the experience as well (Crowe et al., 2023). The authors concluded that patients need choices, information and the chance to decide on treatment in an informed way (Crowe et al., 2023). In the synthesis, 59% of its participants came from a single study of 1747 people, and 70% of participants were women (Crowe et al., 2023).

Types of Major Depressive Disorder

A type of major depressive disorder here means a recognised way of classifying an episode or its course. It does not imply a separate disease. The current diagnostic manual sorts the disorder in three ways: by course (a single episode or recurrent episodes), by current severity and remission status, and by specifiers that describe features of the present episode (American Psychiatric Association, 2022). Other types on this list are research constructs defined by a proposed cause, historical diagnoses that were later absorbed, or neighbouring disorders that share the depressive presentation and sit at its boundary.

The cognitive evidence is uneven across these types. For a few, such as psychotic, melancholic and first-episode depression, meta-analyses compare cognition directly with other forms of the disorder. For most specifiers, no study cited on this page makes that comparison, and each subsection below says so where it applies.

Table 1

Types of Major Depressive Disorder and Their Standing

TypeStandingKey instrument
Single episodeDSM-5-TR course or severity descriptor–
RecurrentDSM-5-TR course or severity descriptor–
Severity: mild, moderate and severeDSM-5-TR course or severity descriptorPHQ-9
With anxious distressDSM-5-TR specifierHamilton anxiety/somatization factor score
With mixed featuresDSM-5-TR specifier–
With melancholic featuresDSM-5-TR specifier–
With atypical featuresDSM-5-TR specifierColumbia criteria
With psychotic featuresDSM-5-TR specifier–
With catatoniaDSM-5-TR specifier–
With peripartum onsetDSM-5-TR specifierEdinburgh Postnatal Depression Scale
With seasonal patternDSM-5-TR specifier–
In partial or full remissionDSM-5-TR course or severity descriptor–
Treatment-resistant depressionResearch constructThase and Rush Staging Model
Late-life and vascular depressionResearch construct–
Hopelessness depressionResearch construct–
Involutional melancholiaHistorical diagnosis–
Persistent depressive disorder and chronic major depressionSeparate DSM-5-TR diagnosis–
Bipolar depressionSeparate DSM-5-TR diagnosis–
Bereavement-related depression and prolonged griefResearch construct (bereavement-related depression)–
Minor and subthreshold depressionResearch construct–
Premenstrual dysphoric disorderSeparate DSM-5-TR diagnosis–
Substance- or medication-induced and medically caused depressionSeparate DSM-5-TR diagnosis–

Note. Key instrument is given only where a study cited for that type names one; an en dash marks the rest. Specifiers can co-occur in the same episode.

Single Episode

The single-episode descriptor applies to a first or only major depressive episode (American Psychiatric Association, 2022). A theoretical review reported that first episodes are typically more strongly linked to major life stress than later recurrences are, while presenting its account only as a preliminary framework for weighing competing models against an inconsistent literature (Monroe & Harkness, 2005).

The experience of a first episode has been studied qualitatively. Interpretative phenomenological analysis of interviews with seven first-episode patients in London described a depletion that was relational, bodily and temporal, a shrinking of the person's lived world, and a sense of existential threat; the authors suggested that therapy explore these themes early (Smith & Rhodes, 2015).

Cognitive Implications

Cognitive problems are already present at the first episode. First-episode depression carries small-to-medium deficits in executive functioning, attention, psychomotor speed, and visual learning and memory, according to a meta-analysis drawing on 15 independent samples with 644 adult patients whose mean age was 39.36 years (Lee et al., 2012). Effect sizes varied with remission status, inpatient status, antidepressant use, age and education. The authors cautioned that the primary studies were small and often failed to report demographic and clinical characteristics, and they proposed the deficits as feasible early markers and treatment targets (Lee et al., 2012).

A second, larger meta-analysis points the same way. With 994 first-episode patients pooled, it found performance below that of healthy controls in most cognitive domains (Ahern & Semkovska, 2017). A separate longitudinal part of the analysis, with 4 studies and 92 patients, compared performance at remission with acute performance: processing speed and shifting improved slightly, but impairment in inhibition and verbal fluency persisted. From this the authors drew two conclusions, that the deficits appear early in the illness and that some of them persist once the episode has lifted (Ahern & Semkovska, 2017). Read beside the earlier pooled estimate, the two meta-analyses agree that impairment is measurable at the first episode.

Recurrent

The recurrent descriptor marks a course with more than one episode. A narrative review weighed two explanations for why recurrence is common: that each episode leaves a scar that raises later risk, and that an underlying vulnerability present all along accounts for it. The review concluded that recurrence mostly reflects that lasting vulnerability, largely genetic in origin, though it did not consider biological measures, and it found virtually no evidence of a psychosocial scar and only slight evidence of a scar on cognitions, in children and adolescents but not adults, so on this review the scar hypothesis remains largely unsupported (Burcusa & Iacono, 2007).

The kindling hypothesis makes a related claim about stress. Because the association with major life stress is stronger for first episodes than for recurrences, later episodes may depend less on external triggers; the review that set out this comparison called kindling the primary conceptual framework for this research while finding the literature on it inconsistent, so its standing remains debated (Monroe & Harkness, 2005). A later systematic review found that having had any prior episode raised the risk of recurrence compared with having had none, but two prior episodes carried no significantly greater risk than one, and childhood maltreatment and residual symptoms were among the clearest predictors (Buckman et al., 2018).

Cognitive Implications

Imaging points to a difference between recurrent and first-episode cases. In the REST-meta-MDD Consortium, a resting-state fMRI study of 848 patients and 794 controls across 17 sites, functional connectivity within the default mode network was reduced in patients overall, and the reduction appeared in recurrent depression but not in first-episode, drug-naive patients (Yan et al., 2019). Medication use was associated with the reduction, illness duration was not, and symptom severity tracked connectivity only in the recurrent group; whether treatment drives the change was left untested (Yan et al., 2019).

A cognitive marker predicts who relapses. In 99 remitted outpatients followed for 18 months, those whose dysfunctional attitudes rose after a sad-mood induction relapsed more often, and a rise of 8 or more points went with a shorter time to relapse (Segal et al., 2006). Patients who had recovered with medication showed more of this cognitive reactivity than those treated with cognitive behavioural therapy, and reactivity predicted relapse regardless of prior treatment type. The authors concluded that reactivity may be related to vulnerability to relapse (Segal et al., 2006).

Severity: Mild, Moderate and Severe

Each current episode is graded as mild, moderate or severe (American Psychiatric Association, 2022). The PHQ-9, the nine-item depression module of the Patient Health Questionnaire, is a common way to put a number on that grade. Its validation in 6,000 patients from 8 primary care and 7 obstetrics-gynaecology clinics proposed cut points of 5, 10, 15 and 20; a score of 10 or more detected major depression with 88% sensitivity and 88% specificity against interviews with 580 patients (Kroenke et al., 2001). Higher scores went with poorer function, more sick days and more health care use, with similar results in both clinic types (Kroenke et al., 2001).

Whether severity changes the value of antidepressant medication is disputed. A patient-level meta-analysis of 6 studies (718 adult outpatients) found that the advantage of medication over placebo grew with baseline severity: it was substantial in very severe depression, but below a Hamilton score of 23 the difference fell under d = 0.20, and the authors judged that the benefit in mild or moderate depression may be minimal or nonexistent, on average (Fournier et al., 2010). Among 2,464 patients whose individual data came from six of the eleven eligible Japanese trials, baseline severity did not change how well the drugs worked (interaction coefficient −0.04, with a 95% confidence interval from −0.16 to 0.08; P = 0.49, non-significant): the authors thought it likely that mild, moderate and severe depression gain a similar benefit (Furukawa et al., 2018). The two analyses used different trial pools, and the question remains open.

Cognitive Implications

Severity relates to some cognitive domains and not others. A meta-analysis of correlations found that more severe depression went with poorer episodic memory, executive function and processing speed, whereas semantic and visuospatial memory showed no significant relationship, and timed and untimed tests behaved alike (McDermott & Ebmeier, 2009). The authors suggested that chance variation and low power within studies may explain the inconsistent results, noting that some domains had few studies and that the cross-sectional data limit causal reading (McDermott & Ebmeier, 2009).

With Anxious Distress

The anxious distress specifier marks an episode accompanied by prominent anxiety (American Psychiatric Association, 2022). In a secondary analysis of 2,876 adult outpatients in the STAR*D trial, patients classed as anxious by a Hamilton anxiety/somatization factor score had lower remission rates, took longer to remit and reported more side effects than nonanxious patients, and they also fared worse at the trial's second treatment level (Fava et al., 2008). The authors concluded that anxious depression carries poorer acute treatment outcomes, a finding drawn from data collected for another purpose (Fava et al., 2008).

The specifier itself has been tested for validity. Among 260 patients with major depressive disorder assessed with semistructured diagnostic interviews in the MIDAS project, anxious distress proved common, and the patients who met it had more anxiety and drug use disorders, higher anxiety scores, and poorer functioning and coping than those who did not (Zimmerman et al., 2019). The authors took these differences as support for the specifier's validity (Zimmerman et al., 2019).

Cognitive Implications

No study cited on this page compares the cognition of anxious and non-anxious depression.

With Mixed Features

The mixed features specifier applies when manic or hypomanic symptoms occur during a depressive episode (American Psychiatric Association, 2022), which places it on the border with bipolar disorder. A meta-analysis of 17 studies (20 samples) put the pooled prevalence of DSM-5 mixed features at 11.6% (95% CI 7.9–16.7%) in major depressive episodes and 26.8% (95% CI 17.0–39.5%) in manic or hypomanic episodes, with the lowest figures for major depressive disorder in East Asian countries (Na et al., 2021). Most included studies carried a moderate to high risk of bias, much of the heterogeneity went unexplained, and the authors called for more prevalence research (Na et al., 2021).

The specifier itself is contested. A historical and systematic review argued that the DSM-5 criteria for mixed features have a weak empirical basis and could harm treatment (Koukopoulos & Sani, 2014).

Cognitive Implications

No study cited on this page reports the cognition of depression with mixed features.

With Melancholic Features

The melancholic features specifier describes a recognised clinical picture within major depression (American Psychiatric Association, 2022). A position paper written for the DSM-5 revision argued that melancholia deserves recognition as a separate mood disorder, on the grounds that it is more biologically homogeneous than major depression and responds to treatment in a more specific way (Parker et al., 2010). The same authors held that the DSM-IV specifier failed to separate melancholia from other depression, partly because several of its features are shared with major depression in general, and that clinical assessment neglected it (Parker et al., 2010).

Cognitive Implications

Melancholic patients show somewhat poorer cognition, with qualifications. A meta-analysis of 9 studies found that melancholic patients performed worse than non-melancholic patients in attention and working memory, visual learning, and reasoning and problem solving, with effect sizes between −0.31 and −0.46 (Zaninotto et al., 2016). In sensitivity analyses, no difference was detected among drug-free patients. Melancholic patients were also older and more severely ill, although neither age, severity nor treatment emerged as a moderator. The authors described the effect as moderate and, seemingly, specific to melancholia, although no difference was detected among drug-free patients (Zaninotto et al., 2016).

With Atypical Features

The atypical features specifier is built around mood reactivity, the capacity of mood to brighten in response to positive events (American Psychiatric Association, 2022; Łojko & Rybakowski, 2017). A narrative review reported that its prevalence varies with the criteria applied and the clinical setting, and that it is associated with a younger age of onset (Łojko & Rybakowski, 2017). On treatment, the review reported an advantage for monoamine oxidase inhibitors over tricyclics (effect size 0.27) but not over SSRIs (0.02), and no subtype difference in response in the iSPOT-D study. The same review stressed that most biological findings overlap with non-atypical depression, that biochemical results are often disputed and that no comprehensive treatment guidelines exist (Łojko & Rybakowski, 2017).

Cognitive Implications

The cognitive evidence is indirect. The same review reported shorter evoked-potential latencies in atypical than in melancholic depression, and no difference between the two in hippocampal volume or white matter (Łojko & Rybakowski, 2017). Direct neuropsychological comparisons of atypical depression are not among the studies cited on this page.

With Psychotic Features

The psychotic features specifier applies when delusions or hallucinations occur during the episode (American Psychiatric Association, 2022).

Cognitive Implications

Psychotic depression carries a clearer cognitive cost than most other types. Across 12 studies, patients with psychotic major depression performed worse overall than those without psychosis (ES = 0.38), and the gap was larger among drug-free patients (ES = 0.69) (Zaninotto et al., 2015). Deficits appeared in verbal and visual learning, processing speed, the Trail Making Test A, and the WAIS-R digit span backward and digit symbol tasks, while the groups did not differ on sociodemographic or clinical variables. The authors judged that a link between psychosis and cognitive deficit seems supported (Zaninotto et al., 2015).

With Catatonia

The catatonia specifier can be added to a major depressive episode (American Psychiatric Association, 2022). Catatonia occurs across many conditions: a meta-analysis of 74 studies (k = 80, n = 110,764) estimated its pooled prevalence at 9.0% (95% CI 6.9–11.7), rising to 23.9% in samples defined by electroconvulsive therapy or high creatine phosphokinase (Solmi et al., 2018). The results were highly heterogeneous (I² = 98%) and showed publication bias (P < .01), so the authors advised cautious interpretation (Solmi et al., 2018).

Cognitive Implications

No study cited on this page reports cognitive findings specific to catatonic depression.

With Peripartum Onset

The peripartum onset specifier applies when an episode begins during pregnancy or after delivery (American Psychiatric Association, 2022). A systematic review with meta-regression of 140 estimates from 96 studies put the pooled prevalence of perinatal depression at 11.9% (95% CI 11.4–12.5) (Woody et al., 2017). Estimates were higher when symptom scales rather than diagnostic instruments were used and higher in low- and middle-income than in high-income countries, suggesting a heavier burden there. The model explained only 31.1% of the variance, data from low-income countries were scarce, and there were too few data to pool incidence (Woody et al., 2017).

Screening shows that not every postpartum depression is unipolar. In a sequential series of 10,000 mothers screened by telephone with the Edinburgh Postnatal Depression Scale (a positive screen being a score of 10 or higher) and then interviewed, most of the screen-positive mothers who were evaluated had unipolar depression, while a minority had bipolar disorder (Wisner et al., 2013). Episode onset fell after delivery in 40.1% of cases, during pregnancy in 33.4% and before pregnancy in 26.5%. Of the screen-positive mothers, 59.2% completed home visits and 10.5% a telephone diagnostic interview, and the diagnoses of those who declined are not known (Wisner et al., 2013).

Cognitive Implications

No study cited on this page reports cognitive findings for depression with peripartum onset.

With Seasonal Pattern

The seasonal pattern specifier describes episodes that recur at a particular time of year (American Psychiatric Association, 2022). The clinical description goes back to a case series of 29 patients whose depressions recurred at the same time each year, most of them with bipolar disorder, especially bipolar II, with hypersomnia and carbohydrate craving among the features studied; a preliminary trial in the same report suggested that bright artificial light may act as an antidepressant (Rosenthal et al., 1984).

On treatment, a meta-analysis of randomized trials in seasonal affective disorder found that bright light therapy lowered depression ratings more than placebo (SMD = –0.37, 95% CI –0.63 to –0.12; 18 studies, 610 patients) and raised response rates (RR = 1.42, 95% CI 1.08–1.85; 16 studies, 559 patients) (Pjrek et al., 2020). Heterogeneity was moderate, risk of bias moderate to high and samples small to medium, so the authors, while regarding light therapy as effective, called for larger high-quality trials; they found no sign of publication bias (Pjrek et al., 2020).

Population data cut the other way. A cross-sectional survey of US adults using the PHQ-8 found no association between depression and latitude, season or sunlight exposure, and its authors suggested the seasonal modifier be reconsidered for discontinuation (Traffanstedt et al., 2016). The two lines of evidence ask different questions, one about treatment response in diagnosed cases and one about symptom levels in the population, and they have not been reconciled.

Cognitive Implications

No study cited on this page reports cognitive findings for the seasonal pattern.

In Partial or Full Remission

Remission is recorded as partial or full once criteria for an episode are no longer met (American Psychiatric Association, 2022). A three-wave population survey of 4,796 Dutch adults (NEMESIS) measured vulnerability traits, including neuroticism, self-esteem, mastery and a depressive coping style, before, during and after an episode (Ormel et al., 2004). People who went on to a first episode were already more vulnerable beforehand; vulnerability rose during the episode and returned to its premorbid level afterward, with no scar effect even after severe or long episodes. The authors cautioned that residual and prodromal symptoms could bias such before-and-after comparisons, and suggested that trait and state effects run as two independent processes (Ormel et al., 2004).

Cognitive Implications

Some deficits outlast the episode. A meta-analysis of studies using the CANTAB battery found that patients in remission still performed below controls in executive function and attention, whereas their memory difference was not significant (Rock et al., 2014). The authors argued that cognitive impairment is a core feature of depression that is not entirely secondary to low mood (Rock et al., 2014).

A much larger synthesis reached a similar verdict. Across 252 studies comparing 11,882 remitted patients with 8,533 healthy controls, 55 of 75 cognitive variables (73%) showed a significant deficit, and most of these were small (30 variables) or medium (22) in size (Semkovska et al., 2019). Remitted patients fell below controls in processing speed, working memory, executive functioning and long-term memory, but not in auditory attention or autobiographical memory, and deficits were greater in those with more previous episodes. The authors concluded that the deficits persist and worsen as episodes accumulate, after noting that earlier results had been inconsistent (Semkovska et al., 2019).

Treatment-Resistant Depression

Treatment-resistant depression is not a DSM category, and it has no agreed definition. A systematic review of 150 studies found 155 different definitions; about half of the definitions required at least two failed treatments, and the Thase and Rush Staging Model was the most used staging model, at 14.2% (Brown et al., 2019). The review covered only studies from 2016 and 2017 and drew on only six key informants, and it called for a definition shared by research and practice (Brown et al., 2019).

A later review likewise found no consensus definition, and its verdicts differed by treatment option: the evidence for extending, switching and combining antidepressants was mixed, psychotherapy alone was not established as efficacious but gave significant relief when added to an antidepressant, and very few factors predicted non-response across treatments (McIntyre et al., 2023). An open-label randomized trial at 5 sites assigned 403 patients with non-psychotic treatment-resistant depression to ketamine or electroconvulsive therapy (ECT). In this open-label comparison ketamine met the noninferiority test against ECT: 55.4% of the ketamine group responded, compared with 41.2% of the ECT group, a gap of 14.2 percentage points whose interval ran from 3.9 to 24.2 (P<0.001) (Anand et al., 2023). Quality of life did not differ; ECT brought more musculoskeletal effects and ketamine more dissociation, and 38 patients withdrew before treatment began (Anand et al., 2023).

Cognitive Implications

Memory differed in that trial: delayed recall on the Hopkins Verbal Learning Test-Revised appeared to drop after 3 weeks of ECT but changed little with ketamine, and the ECT group recovered gradually during follow-up (Anand et al., 2023). Because the trial was open-label, patients and clinicians knew which treatment was given.

Late-Life and Vascular Depression

Under the vascular depression hypothesis, cerebrovascular disease contributes to some depressive syndromes of later life, perhaps by disrupting prefrontal systems once lesions build up past a threshold, and the authors point to the link between ischaemic lesions and distinctive behavioural symptoms as support (Alexopoulos et al., 1997). Its originators offered it as a hypothesis likely to generate useful studies, limited to some late-life syndromes rather than all of them. A later narrative review set out candidate mechanisms (disconnection, inflammation and hypoperfusion) and stressed that these remain hypotheses still to be tested (Taylor et al., 2013).

Cognitive Implications

In the original proposal, disruption of prefrontal systems was the hypothesised route from vascular lesions to depression (Alexopoulos et al., 1997). The later review linked executive dysfunction to poorer antidepressant response, and progression of cerebral hyperintensities or cognitive deficits to a worse course (Taylor et al., 2013). Late-life depression is also associated with later dementia: a meta-analysis of 23 cohort studies (49,612 participants) found pooled risks of 1.85 (95% CI 1.67–2.04) for all-cause dementia, 1.65 (95% CI 1.42–1.92) for Alzheimer's disease and 2.52 (95% CI 1.77–3.59) for vascular dementia, the last significantly higher than the Alzheimer's figure in a subgroup analysis of five studies (P = 0.03) (Diniz et al., 2013). The vascular estimate rested on only five studies, none of the cohorts used structured diagnostic interviews, and the finding is an association; the section Cognitive Impairment and Dementia in Later Life takes it further.

Hopelessness Depression

Hopelessness depression is a theory-defined subtype, identified by its proposed cause rather than by its symptoms. Hopelessness theory, a cognitive diathesis-stress account, holds that a negative attributional style interacting with stress produces hopelessness, which in turn produces this form of depression (Abramson et al., 1989). Its authors called the theory promising but in need of further research, noted that the symptoms of hopelessness depression had been little studied, and allowed that hopelessness might be better understood as a cause of depression than as the mark of a subtype (Abramson et al., 1989). The broader expectancy construct of pessimism has its own page.

Cognitive Implications

The theory's prospective test supports its cognitive premise. Over 2.5 years, 347 college freshmen with no initial disorder who held negative cognitive styles had 3.5 to 6.8 times greater odds than low-risk peers of developing major, minor or hopelessness depression (Alloy et al., 2006). For hopelessness depression, the risk applied equally to first onsets and to recurrences.

Involutional Melancholia

Involutional melancholia was a twentieth-century diagnosis given mostly to older, postmenopausal women. A historical analysis of American psychiatry from 1900 to 1980 traced how the diagnosis was eventually folded into major depression (Hirshbein, 2009).

Cognitive Implications

No study cited on this page examines cognition under the historical diagnosis of involutional melancholia.

Persistent Depressive Disorder and Chronic Major Depression

Persistent depressive disorder is a separate diagnosis that brings together dysthymia and chronic major depression; one review of it emphasised a dimensional perspective based on course (Schramm et al., 2020). A narrative review described it as often more disabling than episodic major depression and noted that dysthymia can be difficult to detect until a major depressive episode is superimposed. Information on its causes is scarce; the authors judged them likely multifactorial and proposed a modular approach to treatment (Schramm et al., 2020).

Cognitive Implications

No study cited on this page compares cognition in persistent depressive disorder with cognition in episodic major depression.

Bipolar Depression

Depressive episodes that occur in bipolar disorder belong to a separate diagnosis, but the boundary is not sharp. Mixed features sit on it: they appeared in 26.8% of manic or hypomanic episodes in one meta-analysis (Na et al., 2021), and critics hold that the DSM-5 criteria for recognising them have a weak basis (Koukopoulos & Sani, 2014). The review of atypical depression also reported a survey in which patients with bipolar I disorder, but not bipolar II, were more likely than those with major depression to display atypical features (Łojko & Rybakowski, 2017).

Cognitive Implications

Cognition does not cleanly separate the two. A meta-analysis of studies that compared unipolar and bipolar patients directly found no difference between them during depressive episodes; in euthymia the only difference was better verbal memory in the unipolar group (g = 0.64, P < 0.001), with the other measures alike (Samamé et al., 2017). The authors concluded that distinct cognitive profiles for the two disorders cannot currently be proposed, and cautioned that differences in sample characteristics and in exposure to adverse factors may explain the gaps that were found (Samamé et al., 2017).

Whether depression after a death should be diagnosed like other depression was disputed during the DSM-5 revision. Normal grief can resemble a depressive episode, and the older manuals excluded bereavement from the diagnosis. Using National Comorbidity Survey data from 8,098 people aged 15 to 54, one study found that uncomplicated depression after bereavement did not differ significantly from uncomplicated depression after other losses on 8 of 9 indicators, and that both differed from complicated cases on most of them (Wakefield et al., 2007). Its authors concluded that the data did not support singling out bereavement alone for exclusion. A narrative review took the opposite route to policy: bereavement-related episodes resemble other episodes in genetics, comorbidity, chronicity and recurrence, and medication response, so the exclusion should be dropped from DSM-5 (Zisook et al., 2012). That review also reported that a personal or family history of depression predicted depression after bereavement. Grief that persists in its own right is classified separately as prolonged grief disorder, a boundary condition covered on its own page.

Cognitive Implications

No study cited on this page compares cognition in bereavement-related and other depression.

Minor and Subthreshold Depression

Depressive symptoms that fall short of the full criteria still carry risk. A meta-analysis of 16 longitudinal cohort studies (n = 67,318) found that subthreshold depression roughly doubled the incidence of later major depression (IRR = 1.95, 95% CI 1.28–2.97), with similar ratios in youth, adults and older people and in community and primary care samples (Lee et al., 2019). The authors read this as favouring indicated prevention in any age group and setting (Lee et al., 2019).

Cognitive Implications

Cognitive style predicts minor depression too. In the prospective study of college freshmen described under hopelessness depression, negative cognitive styles raised the odds of minor depression, and the effect was stronger for first onsets than for recurrences (Alloy et al., 2006).

Premenstrual Dysphoric Disorder

Premenstrual dysphoric disorder is a separate diagnosis within the depressive disorders. A meta-analysis of 44 studies (50,659 participants) estimated its point prevalence at 3.2% (1.7%–5.9%) when the diagnosis was confirmed and 7.7% (5.3%–11.0%) when it was provisional, falling to 1.6% (1.0%–2.5%) in community samples with confirmed diagnosis (Reilly et al., 2024). Heterogeneity was high and few studies applied the full DSM criteria in community settings; the authors concluded that provisional diagnosis likely inflates prevalence (Reilly et al., 2024).

Cognitive Implications

No study cited on this page reports cognitive findings for premenstrual dysphoric disorder.

Substance- or Medication-Induced and Medically Caused Depression

Depression attributable to a substance, a medication or another medical condition is diagnosed separately from major depressive disorder (American Psychiatric Association, 2022). Depression after stroke illustrates the medical case. A meta-analysis of 61 studies (25,488 people) put its pooled frequency at 31% (95% confidence interval 28% to 35%), lower at one to five years (25%) and at five years (23%) after the stroke (Hackett & Pickles, 2014). The figure was not significantly different from the 33% of the authors' 2005 review, though frequency varied considerably between studies (Hackett & Pickles, 2014).

Comorbidity with physical illness also compounds harm. In the World Health Survey of 245,404 adults in 60 countries, depression lowered health scores more than angina, arthritis, asthma or diabetes, and depression combined with a chronic disease produced the worst health of any group (Moussavi et al., 2007). The survey was cross-sectional, so it shows the size of the decrement, not its direction.

Cognitive Implications

No study cited on this page reports cognitive findings specific to substance-induced or medically caused depression.

Prevalence, Course and Burden

Major depressive disorder is among the most common mental disorders, and much of what is known about its frequency comes from large household surveys that interview community residents rather than patients. In the United States, the National Comorbidity Survey Replication interviewed 9090 adults face to face, using the Composite International Diagnostic Interview to assign diagnoses and the Quick Inventory of Depressive Symptomatology Self-Report and the Sheehan Disability Scale to grade severity and role impairment. It estimated lifetime prevalence at 16.2% and 12-month prevalence at 6.6% (Kessler et al., 2003). The disorder was rarely mild in that sample: of the cases, only 10.4% fell in the mild band, against moderate (38.6%), severe (38.0%) and very severe (12.9%) ratings for the rest (Kessler et al., 2003).

The same survey found a wide gap between having the disorder and receiving adequate care. Some treatment was reported by 51.6% of people with 12-month major depressive disorder, but only 41.9% of those treated received care that met the survey's adequacy criteria, and sociodemographic correlates of who was treated were far fewer than correlates of who was ill (Kessler et al., 2003). The authors judged this inadequate treatment a serious concern (Kessler et al., 2003).

The World Mental Health surveys extended the picture to 89,037 adults in 18 countries, assessing major depressive episodes against DSM-IV criteria with the same interview and measuring impairment with the WHO Disability Assessment Schedule-II (Bromet et al., 2011). Lifetime prevalence of an episode was higher in the high-income countries than in the low- and middle-income countries (14.6% against 11.1%), yet 12-month prevalence did not differ significantly between the two groups (5.5% and 5.9%), and the higher ratio of 12-month to lifetime prevalence in the low- and middle-income countries could reflect greater persistence, though the authors judged differential recall error the more parsimonious explanation (Bromet et al., 2011). Average age of onset was similar across income levels, although onset was dated retrospectively (Bromet et al., 2011). Women outnumbered men by roughly two to one, and the marital status most strongly tied to depression varied by setting: being separated in high-income countries, being divorced or widowed in low- and middle-income countries (Bromet et al., 2011). Those odds ratios were bivariate and unadjusted, and the surveys did not test how the risk factors combine (Bromet et al., 2011). People with a current episode reported more impairment than people who had never had one, impairment rose with how recent the episode was, and countries with higher prevalence also tended to report greater impairment (Bromet et al., 2011). The investigators concluded that depression is a major public-health concern everywhere they looked and is closely bound up with social conditions (Bromet et al., 2011).

Course is harder to summarise than frequency. Major depression presents in varied ways, follows a course and prognosis that are hard to predict, and responds unevenly to treatment, and it is severely disabling, with costs to psychosocial functioning and quality of life (Malhi & Mann, 2018).

At the population level, the Global Burden of Disease 2019 analysis modelled 12 mental disorders across 204 countries from 1990 to 2019. Disability-adjusted life-years (DALYs) attributed to mental disorders as a group rose in absolute terms from 80·8 million to 125·3 million, while the age-standardised rate stayed essentially flat (1581·2 and 1566·2 DALYs per 100 000) (GBD 2019 Mental Disorders Collaborators, 2022). The age-standardised rate was higher in females than in males, and nearly all of the burden came from years lived with disability rather than years of life lost (GBD 2019 Mental Disorders Collaborators, 2022). That split is partly an artefact of method: deaths were attributed directly only to eating disorders, so premature mortality linked to depression is not reflected in these figures (GBD 2019 Mental Disorders Collaborators, 2022). The collaborators found no evidence that the global burden of mental disorders had fallen over the three decades (GBD 2019 Mental Disorders Collaborators, 2022). Mortality associated with the disorder is taken up in Suicide Risk, and the sex gap in Sex Differences.

The longer a cohort is followed, the larger the share whose disorder returns. A systematic review of the recurrence of major depressive disorder gave recurrence figures for specialised mental healthcare settings of 60%, 67% and 85% at 5, 10 and 15 years respectively, with 35% at the 15-year point in the general population, and recurrence seemed lower in general-population samples than in specialised mental healthcare (Hardeveld et al., 2010). Clinical factors appeared to matter most: a history of previous episodes and the persistence of subclinical residual symptoms each raised the likelihood of recurrence, whereas gender, civil status and socioeconomic status showed no relation to it (Hardeveld et al., 2010). The same review noted how little data existed from the general population and from primary care (Hardeveld et al., 2010).

Depression also carries excess mortality. Pooling 293 prospective studies with 1,813,733 participants from 35 countries, a meta-analysis found that people with depression died at a higher rate than those without, with relative risks reported between 1.52 and 1.64 (Cuijpers et al., 2014). The association held at a similar size in community and patient samples, with the exception of patients with chronic obstructive pulmonary disease (COPD), in whom the risk was higher; it did not differ between heart disease, cancer and kidney disease samples, between diagnostic interviews and self-report measures, or between world regions (Cuijpers et al., 2014). Larger relative risks came from studies with shorter follow-up and from those at higher risk of bias, and heterogeneity across studies was high; the authors accepted that the excess mortality is real but cautioned that publication bias and weak study quality may have inflated its size (Cuijpers et al., 2014).

Beck's Cognitive Model of Depression

The cognitive model treats depression as, in large part, a disorder of how information about the self, the world and the future is processed. Aaron Beck set out its foundations in his 1967 book on the clinical, experimental and theoretical aspects of depression (Beck, 1967). Its standing differs sharply between its two kinds of claim. As a description of depressive thinking it is well supported; as a causal theory of who becomes depressed it is only partly supported and remains debated.

Components and How They Are Said to Work

The model is a diathesis-stress account built in layers. At its base are dysfunctional beliefs, enduring negative assumptions about the self. In his review of more than 40 years of the model's development, Beck described a developmental sequence: early trauma, combined with dysfunctional beliefs, lays down a predisposition, and later stressors that match the content of those beliefs then precipitate an episode (Beck, 2008). Once activated, the beliefs are said to bias information processing toward the negative. They surface as automatic thoughts, quick negative appraisals that arrive unbidden, and as cognitive distortions, systematic errors in thinking. Beck proposed that the part genetic polymorphisms play in overreaction to stress in the development of depression is probably mediated by cognitive distortions (Beck, 2008).

Put plainly, the model has three layers. Dysfunctional beliefs act as predisposing factors, stressors in later life that are congruent with them act as precipitating factors, and the model also includes automatic thoughts and cognitive distortions (Beck, 2008). The middle step, activation, is the one that later research tried to capture directly. Under the label of cognitive reactivity, studies using priming and longitudinal designs asked whether a vulnerable person's negative constructs become active under stress, a test of the diathesis-stress claim that is reviewed under Standing of the Model below (Scher et al., 2005). Dysfunctional beliefs have been measured with the Dysfunctional Attitudes Scale, one of the instruments that studies of cognitive change in therapy have used (Lorenzo-Luaces et al., 2015).

The same review tied the psychological layers to biology. Beck linked depression to a hyperactive amygdala alongside a hypoactive prefrontal cortex, and treated amygdala hypersensitivity and a genetic polymorphism as risk factors, the polymorphism being associated with overreaction to stress and with hypercortisolemia, a chronically raised level of the stress hormone cortisol (Beck, 2008). He was explicit about the limits of this synthesis. The pathways from genes to beliefs were only sketched, the interaction between them had been demonstrated only recently, and the claim was that integrated psychological and biological study may clarify depression, not that it already had (Beck, 2008).

The Evolutionary Revision

A later theory paper recast the model in evolutionary terms. Beck and Bredemeier proposed that depression is an adaptation for conserving energy after a perceived loss, with anhedonia and anergia among its features (Beck & Bredemeier, 2016). In their integrative model, negative information processing and strong stress reactivity both feed depressogenic beliefs, and those beliefs in turn make appraisals of loss more likely (Beck & Bredemeier, 2016). Severe depression, on this account, is an anachronistic evolutionary program: its symptoms were presumably adaptive in ancestral conditions and are maladaptive now (Beck & Bredemeier, 2016). The paper is a theoretical proposal rather than a test, and the adaptive claim carries the authors' own hedge. Other evolutionary accounts are taken up under Interpersonal and Evolutionary Accounts.

The Neural Account

Disner and colleagues reviewed some 40 years of work and observed that the neural findings had not been integrated with Beck's framework (Disner et al., 2011). Their synthesis associated the model's negative biases, in attention, in memory, in rumination and in dysfunctional attitudes and schemas, with a stronger pull from subcortical emotion regions and weaker top-down cognitive control (Disner et al., 2011). A narrative review spanning 50 years of research pressed the control side of this picture further. In LeMoult and Gotlib's review, depression showed a pull toward negative material and, often, an absence of the positive biases seen in others; the authors judged that impaired cognitive control probably drives these biases, since less reappraisal, more rumination and larger biases in attention and memory went with greater difficulty inhibiting negative content (LeMoult & Gotlib, 2019). These biases are covered in full under Attention, Memory and Executive Biases, and the imaging evidence under Brain Networks and Imaging.

Cognitive Therapy and Cognitive Change

If biased beliefs keep depression going, then procedures that change those beliefs should relieve it, and the change in thinking should come first and carry part of the improvement. That is the logic of cognitive therapy, and Lorenzo-Luaces and colleagues built a conceptual framework for testing it (Lorenzo-Luaces et al., 2015). On their reading of the evidence, the answer was mostly yes on timing: cognitive change tended to be followed by later symptom change, early use of cognitive techniques went with later improvement, and cognitive change was associated with greater resistance to relapse (Lorenzo-Luaces et al., 2015). They concluded that cognitive procedures are effective and that cognitive change contributes to their effect, a claim of contribution rather than sole cause.

Specificity was less clear. Cognitive behavioural therapy and other psychotherapies tended not to differ in how much cognitive change they produced, although in some studies CBT produced more cognitive change than medication (Lorenzo-Luaces et al., 2015). In several studies cognitive change predicted later symptom improvement in cognitive therapy but not under antidepressant medication, though in others cognitive reactivity predicted relapse irrespective of treatment, and evidence on behavioural change as a route to improvement was mixed (Lorenzo-Luaces et al., 2015). The authors were frank about the limits of the causal case. Few studies ran formal mediation tests, cognition and symptoms were often measured at the same time, reverse causation was rarely ruled out, and unmeasured third variables could account for the association (Lorenzo-Luaces et al., 2015). Mechanisms of psychological treatment are treated more broadly under How Psychological Treatments Work.

Standing of the Model

The model's standing has been assessed in reviews that separate what it describes from what it claims to cause. An early literature review by Haaga and colleagues, covering negative self-cognitions, hopelessness, mood-congruent recall and the specificity of loss themes, found the descriptive claims supported but concluded that the causal, diathesis-stress hypotheses lacked strong support (Haaga et al., 1991). Two qualifications came with that verdict. Evidence that depressed thinking is particularly inaccurate or illogical was weak, and fewer studies had tested the causal hypotheses (Haaga et al., 1991).

Later work moved the causal question onto cognitive reactivity, the degree to which a mild sad mood reactivates negative thinking in someone who is not currently depressed. Scher and colleagues reviewed priming and longitudinal studies and judged that cognitive reactivity supports a vulnerability to depression, with the support established in adults and still accruing in children (Scher et al., 2005). The links that matter most for a causal theory, from reactivity to the onset of depression and to relapse, were accruing only slowly, as was the evidence on the attachment origins of cognitive vulnerability (Scher et al., 2005).

A prospective trial then tested reactivity directly. Segal and colleagues followed 99 outpatients aged 18 to 65 whose major depressive disorder had remitted, provoked a sad mood in the laboratory, and tracked relapse over 18 months with the Hamilton Depression Rating Scale and a longitudinal interval follow-up evaluation (Segal et al., 2006). Patients who had recovered on medication showed more reactivity than those treated with cognitive behavioural therapy. Mood-linked reactivity predicted relapse: patients whose dysfunctional attitude scores rose by 8 points or more after the mood provocation relapsed sooner than the others, and this held regardless of the type of earlier treatment (Segal et al., 2006). The authors drew the hedged conclusion that reactivity may relate to vulnerability to relapse. Its use in prevention is discussed under Preventing Relapse and Recurrence.

Taken together, the evidence supports Beck's model as a description of depressive cognition and partly supports it as a causal account. Cognitive reactivity has prospective support as a vulnerability marker for relapse, but whether dysfunctional beliefs cause first onset is still debated, and the biological integration remains, by Beck's own account, a sketch. Whether the biases precede the disorder or follow from it is examined under Cause or Consequence?.

Diathesis and stress in the cognitive model

Switch the latent beliefs on or off and choose the kind of stressor.

Early experienceDysfunctional beliefsLater stressorAutomatic thoughtsDistortionsDepressive symptoms

Beliefs activated: negative automatic thoughts and distortions follow, then symptoms.

Schematic of the diathesis-stress structure Beck described: early experience and dysfunctional beliefs predispose, and a later stressor that matches those beliefs precipitates automatic thoughts, distortions and symptoms (Beck, 2008). It shows the model's claims, not evidence for them; reviews found the descriptive claims better supported than the causal ones (Haaga et al., 1991).

Helplessness, Hopelessness and Rumination

A second family of cognitive accounts places vulnerability less in the content of beliefs about the self (see Beck's Cognitive Model of Depression) than in two habits of mind: how a person explains bad events, and how a person responds once low mood has set in. The first habit is the subject of the attributional reformulation of learned helplessness and of hopelessness theory. The second is the subject of response styles theory and of later work that treats rumination as one kind of repetitive thought.

From Helplessness to Attribution

The reformulation began as a critique. In its original form, the learned helplessness hypothesis linked depression to experience of noncontingency between responses and outcomes, but Abramson and colleagues argued that it could not tell universal helplessness apart from personal helplessness, and that it had nothing to say about whether helplessness was general or confined to one area of life (Abramson et al., 1978). Their remedy drew on attribution theory. On the reformulated account, the explanation a person gives for an uncontrollable outcome influences what they come to expect next (Abramson et al., 1978). Each dimension of that explanation carries a separate prediction: a stable cause should make the helplessness chronic, a global cause should spread it across situations, and an internal cause should lower self-esteem (Abramson et al., 1978). As to standing, the 1978 paper was a theoretical reformulation that set out what it implied for a model of depression; it reported no test of its own (Abramson et al., 1978). Fifty years on, the founders revised the account in the light of neuroscience: passivity, they argued, is the unlearned default response to prolonged aversive events, and what is learned is control, whose detection by the medial prefrontal cortex inhibits the dorsal raphe nucleus, which turns the original theory on its head (Maier & Seligman, 2016). They described what this might mean for treating depression as speculative (Maier & Seligman, 2016).

Hopelessness Theory

Hopelessness theory recast the attributional idea as a cognitive diathesis-stress model and proposed a subtype, hopelessness depression (Abramson et al., 1989). Its attributional diathesis is held to produce depression only when it meets stress, and the authors counted this interaction among the parts of the theory that had been tested (Abramson et al., 1989). Other parts had not. A parallel diathesis, the tendency to infer negative consequences from events, was still untested, as was the claim that the interaction of cognitive style with stress leads to clinical depression (Abramson et al., 1989). The authors listed further open problems: the symptoms of hopelessness depression had been little studied, the stability of the diathesis was unsettled, the idea of specific vulnerability had been examined only in a preliminary way, the theory specified no time lag, and hopelessness might turn out to be a cause of depression rather than the marker of a subtype (Abramson et al., 1989). Its standing in its founding statement was that of a promising theory in need of more research, including tests of what it implies for course, cure and prevention (Abramson et al., 1989). A systematic review of the theory's first quarter century found consistent support for several of its major components, while noting that it is often confused conceptually with the reformulated learned helplessness theory and that clinically significant outcomes have received too little study (Liu et al., 2015).

The strongest test in this group came from a prospective high-risk design. College freshmen with no disorder at the start (N = 347) were divided into high-risk and low-risk groups by their negative cognitive styles and followed for 2.5 years (Alloy et al., 2006). Over that period the high-risk students had greater odds than the low-risk students of major, minor and hopelessness depression, with odds between 3.5 and 6.8 times greater (Alloy et al., 2006). For major depression and for hopelessness depression, cognitive risk predicted recurrences as well as it predicted first onsets; for minor depression, the prediction was stronger for first onsets than for recurrences (Alloy et al., 2006). The study also checked specificity. Cognitive risk did not predict anxiety that occurred without depression, but it did predict the onset of anxiety comorbid with depression (Alloy et al., 2006). The design gives the theory prospective support, because the styles were measured before any episode, though the sample was confined to first-year college students.

Rumination and Repetitive Thought

Response styles theory moves the question from explaining events to responding to mood. Nolen-Hoeksema proposed that ruminative responses to a depressed mood make it last longer, whereas distraction makes it pass sooner (Nolen-Hoeksema, 1991). In the laboratory and field studies her theory paper drew together, people who ruminated stayed depressed for longer than people who distracted themselves (Nolen-Hoeksema, 1991). Its standing in that paper rested on the proposer's own synthesis of these studies rather than on an independent test. A 2020 narrative review treats rumination as a transdiagnostic vulnerability with several proposed sources, executive control and abstract processing among them; the authors offer their model as only a partial answer and regard treatments aimed at these mechanisms as possibly effective on the strength of preliminary trials (Watkins & Roberts, 2020).

An earlier narrative review argued that rumination is better treated as a case of repetitive thought, which can have either constructive or unconstructive consequences (Watkins, 2008). Which outcome follows depends, on this account, on the valence of the thought, the context in which it occurs and its construal level, with abstract repetitive thought linked to more unconstructive outcomes than concrete repetitive thought (Watkins, 2008). The review judged an elaborated control theory account to be the best available framework for these findings, a conclusion that stands as the reviewer's assessment rather than a settled result (Watkins, 2008). Whether attributional style and rumination precede depression or follow from it is a broader question, taken up in Cause or Consequence?.

Interpersonal and Evolutionary Accounts

The cognitive accounts described under Beck's Cognitive Model of Depression and Helplessness, Hopelessness and Rumination place the disorder inside the individual's thinking. A second group of accounts looks outward, to what depressed people do in their relationships, to how other people respond, and to why a response as costly as depression might have been kept by natural selection at all.

Coyne's interactional theory links excessive reassurance seeking both to depression and to rejection by others, and two meta-analyses tested each link (Starr & Davila, 2008). Coyne set out an interactional description of depression in 1976 (Coyne, 1976). Across 38 studies with a combined N = 6,973, reassurance seeking correlated with concurrent depression at r = .32; across 16 studies, its correlation with concurrent interpersonal rejection was weaker, at .14 (Starr & Davila, 2008). The depression association was larger in studies using self-report measures than in studies using interviews and larger in samples with a higher proportion of women, while the gap between community and clinical samples was only marginal (Starr & Davila, 2008). Rejection effects were stronger when the person seeking reassurance reported the rejection than when an informant did, and romantic relationships showed only a marginally stronger effect than nonromantic ones (Starr & Davila, 2008). The authors read the pooled results as support for the model, with several caveats: both pooled effects are concurrent, the prospective studies could be reviewed only qualitatively, and the literature needs more methodological diversity (Starr & Davila, 2008). The interactional model is therefore supported for the cross-sectional associations it predicts, while its causal sequence remains less firmly established.

Beck and Bredemeier proposed an integrative model that sets the cognitive model's constructs inside an evolutionary frame (Beck & Bredemeier, 2016). In their account, negatively biased information processing and strong reactivity to stress each feed depressogenic beliefs, and those beliefs in turn promote appraisals of loss (Beck & Bredemeier, 2016). Depression is then cast as an adaptation that conserves energy once a loss has been perceived, which gives anhedonia and anergia a function rather than treating them only as deficits (Beck & Bredemeier, 2016). Severe depression, on this view, is an anachronistic evolutionary program: the symptoms were presumably adaptive in ancestral conditions but are maladaptive now (Beck & Bredemeier, 2016). The model is a theory paper rather than a test, so its standing is that of a proposed account.

Other evolutionary accounts take a related line. Nesse argued that low mood may be an adaptive response to unpropitious situations, in which pessimism and lack of motivation could be of use, and he presented these ideas as hypotheses consistent with a good deal of evidence but still awaiting specific tests (Nesse, 2000). Andrews and Thomson proposed the analytical rumination hypothesis, on which complex problems trigger depressed affect and that affect in turn fosters rumination and analytic processing of the problem; they judged the hypothesis supported at several levels of evidence while acknowledging that it does not exclude other evolutionary accounts or other hypotheses (Andrews & Thomson, 2009).

The social signal transduction theory of depression connects the interpersonal and biological levels (Slavich & Irwin, 2014). It starts from the observation that social stress, rejection and early life stress all predict depression, and proposes that experiences of social threat raise the production of proinflammatory cytokines (Slavich & Irwin, 2014). A resulting proinflammatory phenotype may, the authors suggest, be a key driver of depression pathogenesis (Slavich & Irwin, 2014). They present the theory as a biologically plausible hypothesis that could offer explanatory and treatment insights, not as an established mechanism, so it too stands as proposed rather than confirmed (Slavich & Irwin, 2014). Its inflammatory arm is taken up further under The Biology of Major Depressive Disorder.

Defeat and entrapment sit at the motivational core of the integrated motivational-volitional model, an ideation-to-action framework for suicidal behaviour (O'Connor & Kirtley, 2018). The model runs from a pre-motivational phase through a motivational phase, in which defeat leads to suicidal ideation by way of entrapment, to a volitional phase, in which separate volitional moderators govern whether ideation becomes an attempt; threat-to-self and motivational moderators act at the earlier transitions (O'Connor & Kirtley, 2018). In the studies O'Connor and Kirtley review, brooding strengthened the step from defeat to entrapment, and people who had attempted suicide differed from those who had only considered it on volitional-phase variables but not, significantly, on motivational ones (O'Connor & Kirtley, 2018). In a cross-sectional self-report survey of 1,288 healthy adults, attempters scored higher than ideators on fearlessness about death, impulsivity and exposure to suicidal behaviour, while the two groups did not differ on discomfort tolerance or on motivational factors (Dhingra et al., 2015). The evidence is not uniform: at a 12-month follow-up, baseline defeat predicted later ideation while baseline entrapment did not reach significance, and thwarted belongingness and perceived burdensomeness did not act as significant moderators (O'Connor & Kirtley, 2018). The authors describe support for the model as growing, but note that prospective studies are few, that defeat and entrapment have mostly been measured at the same time point, and that the temporal order of the factors and their interactions remain unresolved (O'Connor & Kirtley, 2018). The model is taken up further under Suicide Risk.

Accounts that locate depression in relationships also form the theoretical background of interpersonal psychotherapy.

Attention, Memory and Executive Biases

Depression is associated with differences in what a person notices, how ambiguous events are read, what is remembered and how well effortful tasks are performed. The evidence comes mostly from laboratory paradigms that compare depressed and nondepressed groups, and it is strongest for biases in processing emotional material. Whether those biases precede the disorder or follow from it is taken up under Cause or Consequence?, and how well the tasks measure anything stable is taken up under Validity of Laboratory Measures.

Depressed people tend to orient toward negative material, a pattern called attentional bias. Pooling 29 studies of depressed, dysphoric and mood-induced participants, a meta-analysis found a significant negative bias on the dot-probe task (d = 0.52), and its authors read the results as support for biased attention to negative information in depression (Peckham et al., 2010). The bias was clear on the dot-probe task, in which a probe replaces one of two stimuli, but only marginally significant on the emotional Stroop (Peckham et al., 2010). None of the moderators tested (age, sex, sample type, publication year, presentation duration or stimulus type) changed its size, though the authors cautioned that those tests had limited power (Peckham et al., 2010). One review describes depressive cognition as elaborating negative material and disengaging from it poorly, under weak cognitive control (Gotlib & Joormann, 2010). A later review reports that difficulty with inhibition is associated with negative biases in attention (LeMoult & Gotlib, 2019).

A dot-probe trial run

Show the word pair, then reveal the dot and press the side it appears on, as fast as you can. Eight trials.

+

Trial 1 of 8.

Structure after the dot-probe paradigm used in attention-bias research; meta-analysis found a negative attentional bias in depression (Peckham et al., 2010). Eight trials give a noisy score: traditional difference scores of this kind have poor reliability (Rodebaugh et al., 2016), so a single run says nothing about any individual.

Ambiguous information is also resolved in a negative direction. Interpretation bias went with depression at a medium effect size once 87 studies and 9,443 participants were pooled (Hedges' g of 0.72, confidence interval 0.62 to 0.82) (Everaert et al., 2017). Its size was similar in clinical depression, in remitted depression and in people with elevated symptoms, and a negative interpretation bias and a missing positive one contributed about equally (Everaert et al., 2017). Effects were larger when the stimuli referred to the self; instructions to form mental images made no significant difference (Everaert et al., 2017). Method mattered most. Direct measures, which ask people what they think an event means, produced larger effects, whereas indirect measures produced no significant effect, and the authors warned that conclusions depend on how the bias is measured (Everaert et al., 2017).

Memory shows a comparable tilt. An early meta-analytic review examined mood-congruent recall, the tendency to remember affectively toned material that matches one's current mood (Matt et al., 1992). A distinct memory phenomenon concerns personal history. When people recall summarised categories of events rather than specific episodes, the pattern is called overgeneral autobiographical memory, and one theoretical review judged it to be relatively specific to a set of emotional disorders, namely those marked by depression, PTSD or both (Williams et al., 2007). That review built on a hierarchical search model of autobiographical retrieval and proposed three mechanisms that can halt the search at a general level: capture by self-relevant material and rumination, functional avoidance of specific and painful memories, and limits on executive capacity and control (Williams et al., 2007). The elaborated model was offered as a proposed account (Williams et al., 2007).

These biases may share a common root in cognitive control. A narrative review spanning 50 years of research concluded that depression involves negative biases, and often a lack of positive biases, across attention, interpretation and memory, and that deficits in cognitive control probably underlie them (LeMoult & Gotlib, 2019). In that account, difficulty inhibiting negative material goes with more rumination and less use of reappraisal, which may impede the flexible choice of emotion regulation strategies (LeMoult & Gotlib, 2019). An earlier review drew the same connection, linking deficits in inhibition and working memory, rumination and a reduced ability to use positive material to repair mood to depression, and argued that neural and genetic factors would be needed to complete the picture (Gotlib & Joormann, 2010). Direct evidence for the inhibition link comes from a cross-sectional comparison of clinically depressed, formerly depressed and never-depressed adults on a negative affective priming task, where the currently depressed group showed weaker inhibition of negative material than the other two groups (Joormann & Gotlib, 2010). Poorer inhibition of negative material went with more rumination among the currently depressed participants, and over the sample as a whole it was associated with how people regulated emotion, with less frequent reappraisal and more frequent expressive suppression (Joormann & Gotlib, 2010). In the formerly depressed group, reappraisal went with fewer symptoms, while rumination and suppression went with more (Joormann & Gotlib, 2010). The proposed mechanism, then, is that a failure to inhibit negative material ties cognitive control to the emotion regulation strategies a person adopts (Joormann & Gotlib, 2010). Rumination itself is treated more fully under Helplessness, Hopelessness and Rumination.

Beyond biases in emotional processing, depression is associated with poorer performance on neutral cognitive tests. Pooling studies that used the Cambridge Neuropsychological Test Automated Battery (CANTAB), the review found moderate deficits in attention, memory and executive function among currently depressed people relative to controls (Rock et al., 2014). Executive and attentional deficits persisted in remitted patients, whereas the memory deficit was non-significant after remission, and the authors concluded that cognitive impairment is a core feature of depression that is not entirely secondary to low mood (Rock et al., 2014). A broader random-effects meta-analysis of 113 studies compared patients with major depressive disorder against healthy controls and found them impaired across executive function measures, with effects spanning 0.32 to 0.97, as well as slower in processing speed; motor slowing alone was not enough to account for the executive deficits (Snyder, 2013). The deficits were larger with more severe current symptoms and with psychotropic medication, and age showed a weaker association (Snyder, 2013). The author judged the pattern consistent with a theory of broad executive impairment, while noting that the specificity of the deficits needs further research and that a causal link has not been established (Snyder, 2013). Severity matters as well. Correlations pooled across studies linked greater depression severity to worse processing speed, executive function and episodic memory, though not to semantic or visuo-spatial memory, and whether a test was timed made no difference (McDermott & Ebmeier, 2009). Its authors suggested that chance variation and low power within studies may explain the inconsistencies, and they noted that some domains rested on few studies and that cross-sectional designs limit what the correlations can show (McDermott & Ebmeier, 2009).

How large those deficits really are is disputed. In a case-control study of 33 patients with depression and 36 controls, the patients scored lower on an aggregate of cognitive tests by a large margin (d = 0.84) and lower on 6 of 11 individual tests (Moritz et al., 2023). Examiners also rated patients lower on well-being while they were tested; scores on the Impact on Performance Scale, an examiner rating of transient factors like motivation and well-being, predicted how well participants performed and mediated the patient-control difference, and with those factors adjusted the gap fell to a medium size (d = 0.49) (Moritz et al., 2023). The groups did not differ significantly on the motivation subscale itself, and on Matrix Reasoning the adjusted patients outperformed controls (Moritz et al., 2023). The authors concluded that neurocognitive deficits in depression are likely overestimated and recommended reporting the percentage of patients who are impaired rather than relying on group means (Moritz et al., 2023). The study is small and from a single site, most patients were medicated, no structured diagnostic interview was used, and the remaining direct effect was only barely significant (Moritz et al., 2023). It tempers rather than overturns the meta-analytic picture.

A final line of work asks whether depressed people judge some things more accurately than others do. The depressive realism hypothesis holds that nondepressed people show a positive bias that depressed people lack. The hypothesis began with four laboratory experiments with students, in which those who were depressed judged contingency accurately, whereas their nondepressed peers overestimated their control when noncontingent outcomes were frequent or desired (Alloy & Abramson, 1979). A meta-analysis of 75 studies with 7,305 participants from the United States, Canada, Europe and Israel found a small overall depressive realism effect, with nondepressed participants showing more bias than depressed ones (Moore & Fresco, 2012). The effect depended heavily on method: it varied with whether a study had an objective standard of reality and with the paradigm used, and realism effects were larger when symptoms were measured by self-report than by structured interview (Moore & Fresco, 2012). The finding is therefore modest and method-bound, and it does not show that depression confers general accuracy.

Reward Processing and Anhedonia

Anhedonia, a loss of interest in or pleasure from activities that were once enjoyed, is a core symptom of the disorder. It shifts the cognitive question from what a depressed person attends to and remembers to how that person values, learns from and pursues rewards. Laboratory tasks approach this by separating reward processing into components: how strongly an available option is valued, whether choices tilt towards the more rewarded alternative, how quickly behaviour adapts to feedback, and how vigorously a person responds when a reward is on offer.

A systematic review and meta-analysis pooled 48 case-control studies, comparing 1767 people with major depressive disorder against 1387 healthy controls on behavioural reward tasks such as the Cambridge Gambling Task and the Monetary Incentive Delay Task (Halahakoon et al., 2020). Across all tasks the depressed groups performed worse, and the pooled difference was small to medium (SMD = 0.345, 95% CI, 0.209-0.480). By component, the deficit was largest for reward bias, with option valuation and reinforcement learning also impaired, whereas response vigour did not differ significantly between groups (Halahakoon et al., 2020). Several moderator results run against intuition. Wholly unmedicated and at least partly medicated samples showed similar effect sizes, a result the authors judged difficult to interpret, and neither anhedonia scores nor the proportion of women predicted the size of the deficit; samples of older adults did not differ significantly from controls, while studies that matched groups on sex reported larger effects (Halahakoon et al., 2020). The authors suggested that reward research may help characterise the heterogeneity of depression rather than describe every patient.

That evidence carries clear limits. The analysis found significant publication bias overall and significant heterogeneity overall (I² = 68%) and in every domain except option valuation; it sometimes combined dissimilar measures, used effect sizes not adjusted for covariates, and did not examine responses to punishment, reward anticipation or the hedonic experience of reward itself (Halahakoon et al., 2020). Because every study was case-control, the results show that reward processing differs in depression, not that the difference causes the disorder (the question taken up in Cause or Consequence?).

Anticipating Reward and Receiving It

Imaging work fills part of that gap by distinguishing the phase in which a reward is awaited from the phase in which it arrives. One case-control study used functional MRI to compare 30 unmedicated people with major depressive disorder against 31 healthy controls as they performed a monetary incentive delay task (Pizzagalli et al., 2009). When a gain was delivered, the depressed group responded less in the left nucleus accumbens and in the caudate on both sides, and penalty outcomes also drew weaker responses in other parts of the caudate. Group differences while participants were waiting for a possible reward were weaker: the reward cues produced a reduced response only in part of the left posterior putamen. Greater anhedonia and greater depression severity each went with smaller caudate volume. The authors read the pattern as possible basal ganglia dysfunction located mainly in the consummatory phase (Pizzagalli et al., 2009).

Pooled evidence points the same way for outcomes. A meta-analysis combining 38 fMRI studies and 12 EEG studies found that depressed participants showed less striatal activation to reward feedback than healthy participants, together with a reduced feedback-related negativity, an EEG component evoked by feedback (Keren et al., 2018). The reductions in reward anticipation and in the feedback-related negativity were both stronger in samples under age 18, and in longitudinal studies of adolescents lower striatal activation and a blunted feedback-related negativity preceded the onset of depression. The authors proposed that these aberrations may contribute to how depression develops, while noting that the anticipation effect depended on region-of-interest analyses (Keren et al., 2018).

A coordinate-based meta-analysis using activation likelihood estimation adds a second region to the picture. Across 41 whole-brain neuroimaging studies of 794 patients and 803 healthy controls, people with major depressive disorder showed a weaker reward response in the ventral striatum but a stronger one in the right orbitofrontal cortex; an exploratory punishment analysis found heightened response in the left amygdala (Ng et al., 2019). The authors offered dysregulated corticostriatal circuitry as a possible mechanism. Their caveats are substantial: the included studies differed in design and samples, some studies had small samples or reported uncorrected coordinates, most participants were adults in an acute episode, some were taking medication, and publication bias could not be excluded (Ng et al., 2019). The wider imaging literature on depression is treated under Brain Networks and Imaging.

What Anhedonia Means

These results depend on what anhedonia is taken to be. A conceptual review argued that the standard definition of the term is underspecified, because it folds the wanting and pursuit of reward together with the pleasure of having it (Treadway & Zald, 2011). Drawing on dopamine research, the authors held that dopamine bears more on motivation and effort than on consummatory pleasure, and they described a decisional form of anhedonia concerning choices about rewards. They also acknowledged contradictory findings on dopamine and a gap between preclinical models and clinical populations, and concluded that a refined definition of anhedonia may be essential for progress (Treadway & Zald, 2011). The point matters for the task data above: a patient who enjoys a reward when it arrives but will not work for it shows a different deficit from one who finds the reward itself flat.

Reward and the Negative-Bias Accounts

The reward account is distinct from the negative-bias accounts described under Attention, Memory and Executive Biases. Those accounts emphasise an excess of negative processing; the reward account emphasises a shortfall of positive responding. The two are not exclusive. A narrative review of five decades of work concluded that depression involves stronger negative biases, and often a lack of positive biases, across attention, interpretation and memory, and that difficulties in cognitive control probably underlie these biases (LeMoult & Gotlib, 2019). On this reading, the absent tilt towards positive material that healthy people show is itself part of the depressive profile, not merely the shadow of heightened negativity.

Behavioural Activation

Behavioural activation is the psychological treatment most directly aimed at this side of the disorder. In a randomised placebo-controlled trial of 241 adults with major depressive disorder, behavioural activation performed as well as antidepressant medication among the more severely depressed participants, and both outperformed cognitive therapy in that subgroup; initial severity moderated the pattern of outcomes (Dimidjian et al., 2006). The authors discussed implications for treatment guidelines and for dissemination. A comparison of outcomes shows which treatment helped, not how it helped, so whether behavioural activation works by restoring reward processing is a separate question, addressed under How Psychological Treatments Work.

The treatment's rationale is that more activity brings more contact with rewarding experience, so activation and environmental reward are the obvious candidate mediators. A systematic review of 14 prospective treatment studies, which between them tested 21 possible mediators of behavioural activation, found the evidence for both of those candidates weak and reached no firm conclusion about any mediator (Janssen et al., 2021). The reviewers attributed this partly to non-significant results, to poor methodological quality in some studies and to the different questionnaires the studies used to measure processes and beliefs.

Cause or Consequence?

Most of the cognitive features described in the preceding sections were first documented in people who were already depressed. That leaves an awkward question: does a negative way of thinking help to bring an episode on, or does the episode produce it? An early appraisal of the cognitive theory of depression (see Beck's Cognitive Model of Depression) drew the line sharply. Surveying research up to 1991, Haaga and colleagues concluded that the cognitive theory's portrait of how depressed people think had largely held up, whereas its causal hypotheses, including the diathesis-stress ones, still lacked strong support (Haaga et al., 1991). They also noted that fewer studies had tested a causal hypothesis than had described depressed cognition, and that the evidence for depressed thinking being particularly inaccurate or illogical was weak (Haaga et al., 1991). Since then the question has been pursued with three kinds of design: measuring vulnerability before anyone falls ill, provoking it in the laboratory, and changing a cognitive bias experimentally to see whether mood follows.

Prospective high-risk studies establish temporal order. Alloy and colleagues recruited 347 college freshmen with no initial disorder, sorted them into high-risk and low-risk groups by their negative cognitive styles, and followed them for 2.5 years (Alloy et al., 2006). Over that period the high-risk group had 3.5 to 6.8 times greater odds than the low-risk group of major depression, minor depression and hopelessness depression (Alloy et al., 2006). For major depression and for hopelessness depression alike, cognitive risk forecast a recurrence as well as it forecast a first episode; in minor depression, by contrast, it did better at predicting first onsets than recurrences (Alloy et al., 2006). The prediction was also fairly specific: risk status was unrelated to anxiety occurring on its own, and it predicted anxiety only when anxiety arose together with depression (Alloy et al., 2006). Evidence of this kind shows that the styles come first in a student sample; on its own it cannot show that they produce the episodes.

A population cohort can go further, because it observes the same people before, during and after an episode. Ormel and colleagues used three waves of the NEMESIS survey, 4796 Dutch respondents assessed with the Composite International Diagnostic Interview, and tracked four vulnerability measures, neuroticism and self-esteem among them (Ormel et al., 2004). In all, 409 respondents developed a major depressive episode (262 first and 147 recurrent episodes), and 334 were free of an episode again by the third wave (Ormel et al., 2004). Three results followed. People who went on to a first episode were already more vulnerable beforehand than people who never had one; vulnerability rose further while they were ill; and after recovery it fell back to its premorbid level (Ormel et al., 2004). No scar effect appeared, even in the subgroups whose episodes were severe or long-lasting (Ormel et al., 2004). The authors suggested that a persisting trait vulnerability and a temporary state rise operate simultaneously and independently, while cautioning that residual and prodromal symptoms can bias comparisons of before and after (Ormel et al., 2004).

Priming designs address a different problem: a vulnerability may stay silent until something activates it. On this construct-activation view, the measure that matters is cognitive reactivity, the degree to which negative thinking increases when mood is lowered. Scher and colleagues reviewed priming and longitudinal studies and concluded that cognitive reactivity supports a vulnerability account of depression, with the support established in adults and accruing in children (Scher et al., 2005). They were candid that evidence linking reactivity to actual onset and relapse, and to attachment-based origins of the vulnerability, was accumulating only slowly (Scher et al., 2005). Segal and colleagues tested the relapse link in 99 outpatients aged 18 to 65 whose major depressive disorder had remitted during the first phase of a randomised trial, measuring dysfunctional attitudes before and after a sad mood provocation and following patients for 18 months (Segal et al., 2006). Patients who had remitted on medication showed more reactivity than those treated with cognitive behavioural therapy, and patients whose dysfunctional attitudes rose by 8 points or more relapsed sooner than the rest; this held regardless of the treatment received (Segal et al., 2006). The authors put the conclusion cautiously: reactivity may be related to vulnerability to relapse (Segal et al., 2006). How well laboratory measures such as these capture everyday cognition is taken up in Validity of Laboratory Measures.

The most direct causal test changes the cognition and watches the symptoms. Cognitive bias modification (CBM) is a family of training procedures aimed at attention biases, interpretation biases, or approach and avoidance tendencies. A systematic review and network meta-analysis by Fodor and colleagues pooled 85 trials with 5013 adults, but only 20 of those trials (n = 1116) concerned depression; the remaining 65 (n = 3897) concerned anxiety (Fodor et al., 2020). The benefits it found were small and unreliable: prediction intervals were wide and included zero, the depression networks were inconsistent, only four trials were at low risk of bias, and heterogeneity limited confidence throughout (Fodor et al., 2020). Interpretation training outperformed both waitlist and sham training for anxiety, and attention training helped only in post-hoc sensitivity analyses, which excluded trials on post-traumatic stress disorder (Fodor et al., 2020). Those anxiety results do not carry over to depression, for which the review judged the evidence insufficient; even for anxiety its call for larger interpretation-training trials was tentative (Fodor et al., 2020).

Taken together, the evidence runs in both directions and settles neither. Negative cognitive styles precede depressive episodes in a high-risk student cohort, population data show trait vulnerability before onset with no lasting scar afterwards, and reactivity to sad mood predicted earlier relapse in one remitted sample. Each of these designs is observational or correlational at the point that matters, however, and the one family of experiments that manipulates cognition directly has not yet shown that shifting a bias changes depression. A decisive test would need a manipulation that reliably alters the cognitive feature in people at risk, followed by an observed change in later onset or relapse; that study has not been run at the scale the question requires, so whether depressive cognition is a cause, a consequence or a marker of a shared vulnerability remains open. The applied side of the reactivity findings is discussed under Preventing Relapse and Recurrence.

Validity of Laboratory Measures

The cognitive findings described in Attention, Memory and Executive Biases rest on laboratory tasks, and how much those tasks can bear depends on how well they measure. Three problems recur: scores that do not hold steady from one testing to the next, effects that appear on one task and fade on another, and deficits that partly reflect a person's state on the day of testing rather than a stable capacity.

Reliability comes first. An empirical study centred on measurement found that conventional attentional bias scores had poor reliability, whereas dynamic scores performed well (Rodebaugh et al., 2016). The authors' recommendation was that studies report the reliability of their bias indices, and they argued that low values are informative for theory rather than mere noise, partly because the assumption that attentional bias is a stable trait is itself open to question (Rodebaugh et al., 2016). The point matters for programmes such as the Research Domain Criteria and for treatment trials that use bias change as evidence of target engagement, since an unreliable score cannot show either (Rodebaugh et al., 2016). That study concerned attentional bias measurement in general; it did not single out depressed samples.

The task used also shapes the result. A meta-analysis of 29 studies covering depressed, dysphoric and mood-induced participants found evidence of a negative attentional bias, with an effect size of d = 0.52 in the dot-probe studies, which supports the bias account (Peckham et al., 2010). The evidence was not uniform, however: the dot probe showed depressed participants attending more to negative material than nondepressed participants, but the emotional Stroop gave only a marginal difference (Peckham et al., 2010). None of the moderators tested, among them age, sex, sample type, stimulus type, presentation duration and publication year, reached significance, although the authors noted that these tests had limited power (Peckham et al., 2010).

Momentary influences can inflate apparent deficits on standard neuropsychological tests. In a single-site case-control study of 33 patients with depression and 36 controls, patients scored lower on an aggregate of cognitive measures by a large margin (d = 0.84) and on 6 of 11 individual tests (Moritz et al., 2023). Examiners also rated, on the Impact on Performance Scale, how motivation and well-being affected each session; patients were rated lower on well-being during assessment, while the motivation subscale did not differ significantly between groups (Moritz et al., 2023). Ratings on that scale mediated the group difference, and after adjustment the gap fell to a medium size (d = 0.49), with the remaining direct effect only just significant (p = 0.048); on Matrix Reasoning the adjusted scores even favoured the patients (Moritz et al., 2023). The authors concluded that cognitive deficits in depression are likely overestimated and proposed reporting the percentage of patients who are impaired (Moritz et al., 2023). The study's limits are considerable: the sample was small, most patients were taking medication, diagnosis did not use a structured interview and the healthy group may have been unusually high-functioning (Moritz et al., 2023). That last question is the broader one taken up in Cause or Consequence?.

A separate concern is how far laboratory scores speak to life outside the laboratory. Experience sampling methodology offers one complement to task-based testing: a systematic review of 19 studies with clinically depressed individuals, spanning 25 years of work, judged the method informative for depression research across six aspects of the disorder, affect and cortisol secretion among them (Telford et al., 2012). The same review pointed to methodological limitations in the existing studies (Telford et al., 2012). Whether test deficits bear on everyday functioning has also been examined directly. From 488 articles published between 1980 and October 2013, a systematic narrative review retained 10 studies of adults with major depressive disorder under 60 years of age, and a broad association appeared between neurocognitive deficits on neuropsychological tests and impaired psychosocial functioning (Evans et al., 2014). The reviewers rated that evidence weak: the studies were few, carried methodological limitations and differed too much from one another to be pooled in a meta-analysis (Evans et al., 2014).

Stress, Adversity and Onset

Psychological stress does not appear to carry the same weight at every point in the disorder's history. A study of female twins that combined survival and co-twin control analyses, covering 316 onsets, found that stressful life events predicted the onset of major depression, with dependent events carrying more effect than independent ones once level of threat was controlled, and that the odds ratio shrank from the whole sample (OR = 5.64) to dizygotic pairs (OR = 4.52) and again within monozygotic pairs (OR = 3.58) (Kendler et al., 1999). The authors read the association as substantially causal, with roughly a third of it noncausal and reflecting selection (Kendler et al., 1999). Severe life stress shows a stronger association with a first depressive episode than with later recurrences, a pattern behind the kindling hypothesis (Monroe & Harkness, 2005). The kindling hypothesis itself comes from a theory paper by Post, who proposed a stress-sensitisation account in which psychosocial stressors play a larger part in triggering a first episode than in setting off the episodes that follow (Post, 1992). Monroe and Harkness organised the evidence around three major themes and noted that the existing studies were inconsistent. They offered their account only as a preliminary framework for weighing competing models of how stress and recurrence relate, not as an established mechanism (Monroe & Harkness, 2005).

Adversity in childhood bears on the course of the disorder as well as on its onset. A meta-analysis combined 16 epidemiological studies with 23,544 participants and 10 clinical trials with 3,098 participants (Nanni et al., 2012). Adults with a history of childhood maltreatment were more likely to have recurrent or persistent depression than those without one (odds ratio = 2.27, 95% CI = 1.80–2.87). In the trials, maltreated patients were also more likely to show non-response or non-remission (odds ratio = 1.43, 95% CI = 1.11–1.83). Neither association varied significantly with publication bias, the outcome measure, study quality, sample age or the lifetime prevalence of depression in the sample, and the authors concluded that maltreatment predicts an unfavourable course and a poorer treatment outcome (Nanni et al., 2012).

Whether a particular genotype decides how hard adversity strikes has been tested most directly for the 5-HTTLPR polymorphism, and the large pooled test did not bear out a gene-by-environment interaction. The original claim came from a prospective study of a representative birth cohort, in which carriers of the short allele of the serotonin transporter promoter polymorphism who met stressful life events showed more depressive symptoms, more diagnosable depression and more suicidality than long-allele homozygotes, a pattern its authors read as evidence of a gene-by-environment interaction (Caspi et al., 2003). Culverhouse and colleagues ran one uniform R analysis script over 31 datasets and pooled the results by random-effects meta-analysis, with 38,802 European-ancestry subjects in the main analyses (Culverhouse et al., 2018). Exposure to stress strongly raised the odds of lifetime depression (OR 2.16), and men were at lower risk than women (OR 0.57). Carrying the S allele, by contrast, made no significant difference (OR 1.00, p = 0.95), and the allele's interaction with broad stress or with childhood maltreatment was not significant either (maltreatment, lifetime depression: OR 1.05, p = 0.49) (Culverhouse et al., 2018). A handful of nominal results appeared in subsets: a recessive coding of the gene (OR 1.25, p = 0.02), an interaction running in the reverse direction for current depression defined by DSM or ICD criteria, and young-adult analyses that the authors judged fragile because the samples were small (Culverhouse et al., 2018). The pooled test had limits of its own. Some invited groups declined to take part, most of the studies had been designed for other questions, and the analysis could not examine stress lasting under five years, financial stress, persistent or recurrent depression, or non-European samples, which were too small to analyse. Within those bounds, it found no strong interaction between 5-HTTLPR and stress (Culverhouse et al., 2018).

How stress reaches the body is studied along two main routes, taken further in The Biology of Major Depressive Disorder. The older is the hypothalamic-pituitary-adrenal (HPA) axis. Some 40 years of research have tied major depression to hyperactivity of this stress-hormone system, although the mechanisms that produce the hyperactivity are still unclear (Pariante & Lightman, 2008). The second route is immune. People with major depression show higher levels of peripheral inflammatory biomarkers than controls, psychosocial stress increases inflammatory signalling, and patients with high inflammation are more often treatment resistant than those with low inflammation (Miller et al., 2009). Miller and colleagues traced inflammatory cytokines through pathways involving nuclear factor kappa B, neuroendocrine function, neural plasticity and glutamate release and reuptake. They suggested that immune-targeted treatment may help, while stressing that the data on cytokine inhibition were preliminary (Miller et al., 2009).

The social signal transduction theory of depression joins the two strands into a single proposal. On this account, social threat raises proinflammatory cytokine activity and a proinflammatory phenotype, rejection and early-life stress predict depression, and inflammation may be a key driver of depressive illness (Slavich & Irwin, 2014). The authors present it as a biologically plausible hypothesis that could explain why social adversity is depressogenic and might suggest new treatments. Its status is therefore proposed rather than established: the authors put it forward as a biologically plausible theory in which inflammation may be a key driver, not as a demonstrated causal chain (Slavich & Irwin, 2014).

Adversity does not only act on depressed people; on the stress generation account, depressed people may also contribute to the adversity they meet. Hammen followed 60 women in four matched groups for one year, assessing their life events by interview, and found that the women with unipolar depression reported more total stress than normal controls and more interpersonal event stress than any of the other groups (Hammen, 1991). Their excess of dependent events reached only the level of a tendency rather than a firm difference (Hammen, 1991). From this pattern Hammen proposed that generating stress in this way may help to keep a cycle of depression going, a conclusion that rests on a small sample made up entirely of women (Hammen, 1991). A later systematic review of 57 articles judged the evidence that depression is followed by a rise in dependent stress to be substantial and replicated, although its authors called for further research and noted that most studies measured discrete episodic events rather than chronic strain and often relied on self-reported stress (Liu & Alloy, 2010).

The Biology of Major Depressive Disorder

No single biological mechanism accounts for major depressive disorder. A disease primer described the disorder as multifactorial, drawing on findings about the hypothalamic-pituitary-adrenal axis, hippocampal volume and large-scale brain networks, and cautioned that no established mechanism explains every case (Otte et al., 2016). The systems below differ sharply in standing: one long-favoured account is now disputed, several associations are well replicated but small or noisy, and the genetic picture has been redrawn by samples in the hundreds of thousands. Imaging findings are covered under Brain Networks and Imaging, and the biology of sleep under Sleep and Circadian Rhythms.

The serotonin hypothesis, the idea that depression is caused by lowered serotonin (5-HT) activity, is contested. A preregistered systematic umbrella review of 17 reviews and large studies concluded that the evidence offers no consistent support for a lowered-serotonin cause (Moncrieff et al., 2022). In that synthesis, levels of the serotonin metabolite 5-HIAA and of plasma serotonin did not differ between depressed people and controls; evidence of lower 5-HT 1A receptor and serotonin transporter (SERT) binding in depression was weak, inconsistent and confined to some areas; experimentally lowering serotonin through tryptophan depletion had no effect in most healthy volunteers, with only weak evidence of an effect in people whose relatives had depression; and neither the SERT gene nor its interaction with stress predicted depression (Moncrieff et al., 2022). The same review linked antidepressant use to lower serotonin, and its authors acknowledged that the included reviews varied in quality, that earlier antidepressant use was not reliably excluded, and that no review of tryptophan depletion had appeared since 2007 (Moncrieff et al., 2022).

A critical commentary judged that conclusion overstated. Its authors argued that tryptophan depletion affects unmedicated patients with major depressive disorder more than healthy people, that plasma tryptophan is lower in the disorder, and that positron emission tomography (PET) studies point to reduced SERT binding and reduced 5-HT release, so that a role for serotonin, and some perturbation of the system, remains indicated (Jauhar et al., 2023). The commentary conceded weaknesses in its own evidence as well: several of the depletion studies it relied on came from one group, confidence intervals were wide, and removing a potential outlier shrank the effect (Jauhar et al., 2023). The two papers agree that a simple low-serotonin deficit is not demonstrated; they disagree on whether serotonin is implicated at all, and that dispute is unresolved. Neither speaks directly to whether serotonergic drugs work, a question taken up under Treatment Evidence.

The stress-hormone system has a steadier record. A review of roughly 40 years of research concluded that hyperactivity of the hypothalamic-pituitary-adrenal (HPA) axis, the cascade that releases cortisol in response to stress, is associated with major depression, while noting that the mechanisms behind that hyperactivity remain unclear (Pariante & Lightman, 2008). Pooling 671 effect sizes from 361 studies covering 18,454 individuals, a later meta-analysis found that depressed people had higher cortisol and higher adrenocorticotropic hormone (ACTH) than nondepressed people, whereas the difference in corticotropin-releasing hormone (CRH) was not significant (Stetler & Miller, 2011). How large the HPA difference was depended on who was studied. The cortisol gap was wider in older, hospitalised samples than in younger outpatients and wider where melancholic or psychotic features were present, it did not differ significantly between men and women, and methodological differences between studies also contributed to the variation (Stetler & Miller, 2011). The association does not by itself settle whether HPA changes precede the disorder or follow it; the section Stress, Adversity and Onset deals with the environmental side of that question.

Inflammation is the most actively studied newer account. A meta-analysis of 24 studies found higher circulating levels of two proinflammatory cytokines in depressed people than in nondepressed controls: tumour necrosis factor alpha (TNF-alpha), by a weighted mean difference of 3.97 pg/mL (95% CI 2.24 to 5.71; 438 depressed, 350 nondepressed), and interleukin-6 (IL-6), by 1.78 pg/mL (95% CI 1.23 to 2.33; 492 depressed, 400 nondepressed), both at p < .00001 (Dowlati et al., 2010). Six other cytokines, among them IL-1beta, IL-10 and interferon-gamma, showed no significant difference, and individual studies had produced mixed results; the authors read the pooled pattern as strengthening the case that depression involves activation of the inflammatory response system (Dowlati et al., 2010).

Elevated inflammation characterises a sizeable minority of patients rather than all of them. Of 37 included studies covering 13 541 depressed patients and 155 728 controls, a meta-analysis of 30 estimated that 27% (95% CI 21–34%) of patients had low-grade inflammation defined as C-reactive protein (CRP) above 3 mg/L, and 58% (95% CI 47–69%) had CRP above 1 mg/L (Osimo et al., 2019). Compared with controls, patients had higher odds of CRP above 3 mg/L (OR 1.46, 95% CI 1.22–1.75) and above 1 mg/L (OR 1.47, 95% CI 1.18–1.82); the prevalence was not significantly related to sample source, antidepressant treatment, age, body mass index or ethnicity, although it varied widely between studies (Osimo et al., 2019). In the authors' judgement, inflammation may matter for a sizeable proportion of patients, a claim about how common it is, not about whether it causes depression (Osimo et al., 2019).

A review proposed how cytokines might enter the pathophysiology. Inflammatory signalling, which psychosocial stress can increase, was described as interacting with neuroendocrine function, neural plasticity and glutamate release and reuptake, and higher inflammatory biomarkers were associated with resistance to treatment (Miller et al., 2009). The authors suggested that therapies targeting the immune system may help some patients, but the data on blocking cytokines were preliminary (Miller et al., 2009).

The neurotrophin hypothesis holds that depression involves reduced brain-derived neurotrophic factor (BDNF), a protein that supports neuronal growth and plasticity. A systematic review with meta-analyses of 2384 depressed people and 2982 controls found lower serum BDNF in untreated major depressive disorder than in controls, and lower levels in untreated than in treated patients (Molendijk et al., 2014). That review also documented the evidence thinning out. Effect sizes were smaller in more recent studies, smaller in larger studies, and smaller again after adjustment for publication bias; results varied between studies, and the relation between BDNF and symptom severity was inconsistent (Molendijk et al., 2014). The authors judged that the findings still support BDNF as a peripheral marker, but on slimmer evidence than earlier reports implied and with a good deal of noise (Molendijk et al., 2014).

Genetic influence on major depressive disorder is real but modest and spread across many variants. A meta-analysis of 5 family studies and 5 twin studies found that relatives of affected probands had raised odds of the disorder (odds ratio = 2.84, 95% CI = 2.31–3.49), and the twin data put the heritability of liability at 37% (95% CI = 31%–42%), shared family environment at 0% (95% CI = 0%–5%) and individual-specific environment at 63% (95% CI = 58%–67%) (Sullivan et al., 2000). Familial clustering was therefore attributed mainly to genes rather than to a shared upbringing, and it was stronger for recurrent depression; the authors flagged that no adoption study met their criteria and that measurement error is folded into the individual-specific estimate (Sullivan et al., 2000). The heritability figure describes variation in a population, not the share of any one person's illness that is inherited.

Genome-wide studies have since located specific variants. A genome-wide association meta-analysis of 807,553 individuals (246,363 cases and 561,190 controls) identified 102 independent variants, 269 genes and 15 gene-sets associated with depression, with enrichment in prefrontal brain regions; 87 of the 102 variants replicated in a further sample of 1,306,354 individuals (Howard et al., 2019). The same study's Mendelian randomization analyses suggested that neuroticism may raise the risk of depression and that depression may in turn raise the likelihood of having smoked; the opposite direction, from depression to neuroticism, grew weaker and was no longer significant after the variants the two sets of instruments shared were excluded (Howard et al., 2019). The authors presented the results as advancing understanding of the disorder's genetic architecture and as opening avenues for understanding its causes and developing new treatments (Howard et al., 2019).

Those large samples also overturned the candidate-gene era. A preregistered analysis of 18 candidate genes, each studied 10 or more times in earlier work, tested them in subsamples of 62,138 to 443,264 people and found no support for the candidate polymorphisms, for gene-level effects, or for their proposed interactions with environmental exposures such as childhood abuse and socioeconomic adversity (Border et al., 2019). Compared with noncandidate genes, the candidate genes showed no stronger link with depression, and error in measuring the depression phenotype was unlikely to account for this; the authors judged the early hypotheses mistaken and the original positive results probably false positives (Border et al., 2019).

Sleep and Circadian Rhythms

Disturbed sleep is among the symptoms that define an episode: the MeSH definition of the disorder, which credits DSM-5, lists insomnia or hypersomnia nearly every day among them (MeSH, D003865).

Recorded overnight, the sleep of depressed patients looks different from that of controls. A narrative review of the polysomnographic literature by Riemann and colleagues characterised depression by broken sleep continuity and by a disinhibition of REM sleep: relative to controls, patients reached their first REM period sooner, spent more total time in REM sleep and showed higher REM density, while their slow-wave sleep was reduced (Riemann et al., 2020). Suppression of REM sleep appeared to go with treatment response, but the hope that REM abnormalities would mark out subtypes of depression was not borne out (Riemann et al., 2020). Nor is this profile unique to depression. A meta-analysis of controlled polysomnographic studies across several mental disorders, pooling only parameters reported by at least three studies, found sleep continuity worse than in controls in most disorders, and affective, anxiety, autism and schizophrenia groups also differed from controls in sleep depth and REM measures; groups with ADHD or seasonal affective disorder did not differ significantly from controls (Baglioni et al., 2016). No single parameter belonged to one disorder alone, and comorbid groups showed greater REM pressure and more inhibited sleep depth than groups without comorbidity, findings the authors linked to psychiatric comorbidity processes, while reading the sleep continuity disturbances as a transdiagnostic imbalance in the arousal system (Baglioni et al., 2016).

Sleep problems are not only a feature of the episode; they can come before it. Baglioni and colleagues pooled twenty-one longitudinal epidemiological studies published between 1980 and 2010 and found that people who reported insomnia at baseline were more likely than people without sleep difficulties to develop depression later, with pooled estimates of 2.60 (CI: 1.98-3.42) and 2.10 (CI: 1.86-2.38) (Baglioni et al., 2011). The authors read this as a reason to think that treating insomnia early might help prevent depression, a suggestion rather than a demonstrated effect, and two limits temper it: the studies were heterogeneous until outliers were adjusted for, and not all of them took intervening variables into account (Baglioni et al., 2011). The association therefore marks insomnia as a predictor of onset without showing that it causes the later episode, a distinction taken up under Cause or Consequence?.

Trials of insomnia treatment bear on that suggestion only in part. Hertenstein and colleagues meta-analysed 22 randomised controlled trials of cognitive behavioural therapy for insomnia (CBT-I) in people whose insomnia accompanied another mental disorder, with placebo, waitlist or usual care as the comparison (Hertenstein et al., 2022). On self-rating questionnaires CBT-I lowered insomnia severity, though the effect in depression (0.5) was smaller than in post-traumatic stress disorder (1.5); depressive symptoms also improved after treatment (0.5), yet at follow-up the effects on comorbid symptoms were no longer significant (Hertenstein et al., 2022). The authors judged CBT-I effective and worth considering as a first-line treatment for comorbid insomnia (Hertenstein et al., 2022). Easing insomnia in people who already have a disorder is a different matter from preventing depression in people who do not yet have it, and whether insomnia treatment achieves the second remains an open question (Riemann et al., 2020).

Deliberately withholding sleep, one of the chronotherapies applied to depressive episodes, has fared poorly as an add-on. A systematic review and meta-analysis of 29 randomised trials with 1246 participants in unipolar or bipolar major depressive episodes compared interventions given with and without sleep deprivation and found no significant difference in symptom change or in discontinuation (Mitter et al., 2022). The review concluded that there was no evidence that sleep deprivation improves depressive outcomes when added in this way, while noting that the evidence base was conflicting and that adverse events had not been reported (Mitter et al., 2022).

Timing matters as well as quantity. Depression that recurs with the seasons, described under the seasonal-pattern type in Types of Major Depressive Disorder, has been treated with bright light, and a meta-analysis of 19 randomised controlled trials in adults with seasonal affective disorder tested whether it helps (Pjrek et al., 2020). Active light of at least 1,000 lx, delivered by a light box or light visor, was compared with placebo conditions such as dim light of 400 lx or less or negative ion generators. Across 18 studies with 610 patients, depression ratings fell more with bright light than with placebo (SMD = –0.37, 95% CI: –0.63 to –0.12), and across 16 studies with 559 patients the response rate was higher (RR = 1.42, 95% CI: 1.08–1.85) (Pjrek et al., 2020). The review judged bright light therapy effective for seasonal affective disorder while calling the evidence limited: heterogeneity was moderate, the risk of bias moderate to high and the trials mostly small to medium in size, although no sign of publication bias emerged (Pjrek et al., 2020).

Brain Networks and Imaging

Neuroimaging research on major depressive disorder has looked for the circuits behind its cognitive features, and its best-replicated claims concern large-scale networks rather than single regions. The default-mode network (DMN), is the network most often implicated. A meta-analysis paired with a theoretical model reported that patients showed stronger functional connectivity between the DMN and the subgenual prefrontal cortex (sgPFC) than controls, and higher sgPFC blood flow, while DMN blood flow itself did not reliably differ between the groups (Hamilton et al., 2015). The same work found that this connectivity often tracked depressive rumination, and its authors proposed, as a model rather than a tested mechanism, that integration of the sgPFC with the DMN suits the repetitive, self-focused thinking of rumination (Hamilton et al., 2015); the cognitive account of rumination is set out under Helplessness, Hopelessness and Rumination.

Abnormal connectivity in depression does not run in a single direction. Combining 25 resting-state publications, with 556 individuals with MDD and 518 healthy controls, through multilevel kernel density analysis, one meta-analysis found hyperconnectivity within the default network and between the frontoparietal and default networks in patients (Kaiser et al., 2015a). The same patients showed hypoconnectivity elsewhere: within the frontoparietal network, between it and the dorsal attention network, and between salience/emotion and midline systems (Kaiser et al., 2015a). Its authors noted that the studies they pooled varied in both methods and results (Kaiser et al., 2015a).

Larger samples complicated the picture. In the REST-meta-MDD Consortium, 25 research groups pooled resting-state fMRI from 1,300 patients and 1,128 controls, and the sample analysed after exclusions, 848 patients and 794 controls from 17 sites, showed reduced, not increased, DMN connectivity in depression, a reduction found only in recurrent depression and not in first-episode drug-naïve patients (Yan et al., 2019). The reduction was confined to recurrent depression: patients in a first, drug-naive episode did not differ from controls. Medication use was associated with lower DMN connectivity, illness duration was not significantly related to it, and symptom severity showed a positive relation only within the recurrent group (Yan et al., 2019). The consortium kept the DMN central to depression but left untested whether treatment effects run through it, and its analyses of other networks were exploratory (Yan et al., 2019).

A broader neural account links these circuits to the information-processing biases described under Attention, Memory and Executive Biases. A review spanning roughly 40 years of work related negative biases in attention, memory, rumination and dysfunctional schemas to stronger input from subcortical emotion-processing regions together with weaker top-down cognitive control (Disner et al., 2011). The review set out to fill a gap: the neural results had not yet been linked to Beck's cognitive model, the subject of the section Beck's Cognitive Model of Depression (Disner et al., 2011).

The strongest challenge to this literature concerns effect size. In a cross-sectional case-control study from the Marburg-Münster Affective Disorders Cohort Study, 1809 adults aged 18 to 65 (861 with major depressive disorder, 948 controls) completed structural MRI, diffusion-tensor imaging, task-based functional MRI and resting-state MRI (Winter et al., 2022). Patients did differ from controls, but every univariate difference was tiny: the largest in each modality had a partial η2 between 0.004 and 0.017, and the two groups' distributions overlapped heavily. Brain measures classified individuals weakly, about as well as a polygenic risk score and worse than environmental variables, and acute and chronic patients showed the same pattern (Winter et al., 2022). The authors concluded that patients and controls are largely similar at the brain level and that predicting an individual's diagnosis from these measures is not possible, though their estimates are upper bounds for univariate effects only (Winter et al., 2022).

Proposed imaging subtypes have fared no better so far. A replication tested whether resting-state connectivity sorts patients into distinct biotypes of depression using canonical correlation analysis followed by hierarchical clustering. A methodological replication in 187 patients with depression or anxiety from the NESDA and MOTAR samples applied the same steps with permutation tests and found no significant support: connectivity-symptom correlations that looked strong within the sample fell to about zero under cross-validation, and the clusters were no better separated than chance (Dinga et al., 2019). The test was not a direct replication, since it used a different clinical instrument (the Inventory of Depressive Symptomatology), a different preprocessing pipeline and a heterogeneous sample with site effects that limited power, so its authors called for cautious interpretation of the biotype evidence rather than declaring it refuted (Dinga et al., 2019). Together these studies suggest that group-level network differences are real but small, and that they do not yet support diagnosis or subtyping of individual patients; the wider question of heterogeneity is taken up under Heterogeneity and Symptom Networks.

Heterogeneity and Symptom Networks

Two people can both meet criteria for major depressive disorder while having few symptoms in common. The definition is polythetic: it can be satisfied by different subsets of its criteria, with no single profile required. Zimmerman and colleagues measured how much variety that allows in practice. Using semi-structured interviews with more than 1,500 patients diagnosed with DSM-IV major depressive disorder, they found that 170 of the 227 combinations the criteria permit actually appeared, so only 57 went unobserved, and that nine combinations, each seen in more than 2% of patients, together made up more than 40% of diagnoses and could serve as diagnostic prototypes (Zimmerman et al., 2015). The authors judged the heterogeneity large. Their figures describe a clinical sample assessed against DSM-IV criteria, not the general population.

Heterogeneity matters most if the symptoms are not interchangeable, and a narrative review by Fried and Nesse argued that they are not. In the studies they gathered, sad mood explained more of the variance in impairment than hypersomnia did, and hopelessness predicted suicide better than a total score (Fried & Nesse, 2015). Insomnia showed a stronger link to suicidality than hypersomnia, which had no significant association with suicidal behaviour, while insomnia, anxiety and loss of interest each went with poorer treatment response or lower remission (Fried & Nesse, 2015). The kind of life event a person experienced was also tied to which symptoms followed. The instruments compound the problem: the review cited an earlier count of some 280 different depression scales, and the Hamilton Rating Scale for Depression gives somatic items more weight than the Beck Depression Inventory, so totals from different scales are not built from the same symptoms (Fried & Nesse, 2015).

From this, Fried and Nesse concluded that sum scores hide information and that research should analyse symptoms individually, study the causal links among them and use them to personalise treatment (Fried & Nesse, 2015). The network approach to psychopathology treats a mental disorder as a system of symptoms that act causally on one another, including through feedback loops, rather than as the effect of a single common underlying factor (Borsboom & Cramer, 2013). Forbes and colleagues tested how robust such models are by fitting conditional independence networks of major depression and generalized anxiety symptoms in two large epidemiological samples (n = 9282 and n = 8841); the networks replicated poorly, their edge and centrality estimates were both unstable, and the authors named measurement error in single-symptom assessment as one likely reason for that instability (Forbes et al., 2017). The network researchers contested this in a reanalysis of the same two datasets, arguing that the networks held up well both across samples and in split-half comparisons and rejecting Forbes and colleagues' conclusion that they are unreliable; in their account the gap between the two analyses comes from contestable choices of replicability metric and from the interview's skip structure, which they argued distorts the correlations among symptoms (Borsboom et al., 2017). Fried and Nesse's evidence for such links was directional in places: sleep quality predicted next-day affect, whereas daytime affect did not significantly predict that night's sleep, and people deprived of sleep reported lower mood than controls (Fried & Nesse, 2015). The authors were explicit about the costs. The authors also gave reasons for caution: some of the findings they reviewed, such as those on symptoms and impaired functioning, still need replication, analysing depression symptom by symptom is more complicated, the scales in use were often not built or validated with single symptoms in mind, and sum scores keep general information while being less vulnerable to measurement problems (Fried & Nesse, 2015). The network approach is therefore a research programme with open questions, not a settled replacement for the total score.

The criteria list is itself a narrow sample of the syndrome. In a historical analysis of psychiatric textbooks published between 1900 and 1960, Kendler found that the older texts gave more emphasis to cognitive, physical and psychomotor features and described low mood more broadly, whereas the DSM-5 criteria lean more heavily on neurovegetative symptoms (Kendler, 2016). Some of the features the criteria leave out are subtle and hard to assess. Kendler argued that DSM criteria should serve to index a disorder rather than to constitute it (Kendler, 2016). On that view, the heterogeneity counted from DSM symptoms is variety within an already shortened list (see Diagnosing and Measuring Major Depressive Disorder).

A separate question is whether depression is a category at all. Taxometric methods test whether a latent variable behaves like a discrete class (a taxon) or a continuous dimension. A quantitative review of 177 articles reporting 311 distinct findings, with 533,377 participants in total, concluded that most latent variables in psychopathology, mood disorders among them, are dimensional (Haslam et al., 2012). Taxonic findings were less common in recent studies, in methodologically stronger studies and in studies using comparison-data fit indices, and the authors judged that many influential taxonic findings of early research were likely spurious (Haslam et al., 2012). This review pooled many conditions rather than examining depression alone; where it flagged the evidence as not definitive, the areas were schizotypy, autism and substance use. It supports treating depressive severity as a continuum without saying which dimensions should replace the category.

Attempts to cut the heterogeneity along biological lines have not yet held up. One line of work proposed that resting-state fMRI connectivity could sort patients into biotypes of depression, using canonical correlation analysis to link connectivity with symptoms and hierarchical clustering to form groups. In the originating multisite study, functional magnetic resonance imaging in a sample of 1,188 participants yielded four biotypes that differed in limbic and frontostriatal connectivity and could not be told apart on clinical features alone, and its authors proposed that such subtypes may help predict who will respond to neurostimulation, transcranial magnetic stimulation among them (Drysdale et al., 2017). Dinga and colleagues re-tested that approach in 187 patients with depression or anxiety from the NESDA and MOTAR samples (Dinga et al., 2019). Within the sample the canonical correlations were very high (r = 0.99 and r = 0.97), but under cross-validation the association between connectivity and symptoms fell to roughly zero. The clusters' quality indices (CH = 109, silhouette = 0.34) were not statistically significant against simulated data containing no clusters, and the canonical correlations, significant under a chi-square approximation, were not significant under permutation tests (Dinga et al., 2019). The study was a methodological replication rather than a direct one: it measured symptoms with the Inventory of Depressive Symptomatology, used a different preprocessing pipeline, drew on a heterogeneous sample and lost power to site effects, and its authors called for cautious interpretation of the biotype evidence (Dinga et al., 2019). Brain Networks and Imaging covers the wider imaging literature.

Taken together, these studies point the same way without settling the matter. The diagnosis groups very different symptom profiles, the symptoms differ in what predicts them and what they predict, the underlying structure looks more dimensional than categorical, and one prominent biological subtyping did not replicate in an independent, if imperfectly matched, test. Symptom-level and dimensional analyses are the alternatives these sources support, and each is still being tested.

Culture and Social Context

Major depressive episodes occur in every region surveyed, and their distribution tracks social circumstance. The largest cross-national evidence comes from the World Mental Health household surveys, which interviewed 89,037 adults in 18 countries with the Composite International Diagnostic Interview against DSM-IV criteria (Bromet et al., 2011). Lifetime prevalence was higher in the high-income countries than in the low- and middle-income group (14.6% against 11.1%), yet the 12-month figures, 5.5% and 5.9%, did not differ significantly between the two groups (t = 1.2, P = 0.25), and the ratio of current to lifetime episodes was lower in the high-income countries (Bromet et al., 2011). Median age of onset was similar across income groups, in the mid-twenties (25.7 and 24.0 years), although onset was dated retrospectively (Bromet et al., 2011).

The social correlates differed by setting. Women were affected roughly twice as often as men, and younger age went with a higher risk in the high-income countries while older age did so in some low- and middle-income ones (Bromet et al., 2011). Marital disruption showed the strongest link, with separation leading in the high-income countries and divorce or widowhood in the low- and middle-income group; these were bivariate, unadjusted odds ratios, and the combined effect of the risk factors was not tested (Bromet et al., 2011). Impairment, measured with the WHO Disability Assessment Schedule-II, was greater in people with a current episode and rose with recency; Japan's impairment was unusually low and South Africa's was not available (Bromet et al., 2011). The authors judged the disorder a major public-health concern closely bound to social conditions (see Prevalence, Course and Burden for the wider epidemiology).

Socioeconomic position carries its own gradient. Across 51 prevalence studies, Lorant and colleagues' random-effects meta-analysis put the odds of depression higher for people at the lowest socioeconomic level than for those at the highest (odds ratio = 1.81, p < 0.001), with risk declining stepwise as income and education increased (Lorant et al., 2003). The gap was wider for staying depressed than for becoming depressed: low status raised the odds of a new episode modestly (odds ratio = 1.24, p = 0.004) but the odds of a persisting one roughly twofold (odds ratio = 2.06, p < 0.001) (Lorant et al., 2003). The authors called the evidence for this inequality compelling while noting that its size varied widely between studies, depending on region, period, how status was defined and how depression was measured (Lorant et al., 2003). Taking in 289 articles on mental disorders in general, not depression in particular, a later review of reviews grouped social determinants into demographic, economic, neighbourhood, environmental-event and social-and-cultural domains, linked each domain to the Sustainable Development Goals, and suggested that work toward those goals could act on the determinants (Lund et al., 2018).

Culture also shapes how depression is expressed and named. In a clinical and anthropological field study of 100 outpatients diagnosed with neurasthenia in Hunan, China, Kleinman applied DSM-III criteria and found that eighty-seven of them met criteria for major depressive disorder, and read neurasthenia as a form of somatization, distress, most often depression, experienced and presented through bodily complaints (Kleinman, 1982). At follow-up, symptoms improved more than help-seeking declined or than patients' sense of social impairment eased, a judgement of improvement made from the researcher's standpoint rather than the patients' (Kleinman, 1982). Kleinman concluded that neurasthenia can be read in several ways but is most useful clinically when regarded as a biocultural form of somatization (Kleinman, 1982).

A quarter-century later, a direct comparison tested the contrast in both directions. Ryder and colleagues assessed 175 Chinese and 107 Euro-Canadian outpatients by three methods: a spontaneous report of their problems, a structured clinical interview and a symptom questionnaire (Ryder et al., 2008). The Chinese patients reported more somatic symptoms on two of the three, and the Euro-Canadian patients reported more psychological symptoms on all three; the psychological differences were the larger and the somatic ones the less consistent (Ryder et al., 2008). The somatic difference was mediated by externally oriented thinking, a focus on outward events rather than inner states, whereas difficulty identifying or describing emotions did not differ between the groups (Ryder et al., 2008). On that evidence the authors proposed that the Western habit of psychologization, presenting distress in emotional terms, may be the more culture-specific pattern (Ryder et al., 2008).

Local idioms of distress raise a related problem of translation. A systematic review of 138 publications from 1979 to 2014, searched across eight databases, examined thinking too much, an idiom of distress that refers to ruminative, intrusive and anxious thoughts (Kaiser et al., 2015b). Its overlap with depression and other psychiatric diagnoses proved variable, and the idiom conveyed aspects of experience and social position that diagnostic categories did not, so the reviewers concluded that it should be treated neither as a stand-in for any one disorder nor as a single uniform concept (Kaiser et al., 2015b). Its meaning varied widely even within one culture, and only English-language publications were included (Kaiser et al., 2015b).

First-person accounts from young people point in a related direction. A qualitative systematic review with thematic synthesis pooled 39 studies covering 884 adolescents and young people aged 10 to 24 years from 16 countries; sadness, social withdrawal and loneliness were the experiences participants described most often (Viduani et al., 2024). The reviewers suggested that features adolescents themselves report could improve detection, but most of the included studies came from high-income countries and most participants were female (Viduani et al., 2024) (see The Experience of Major Depressive Disorder).

Social threat has also been proposed as a biological pathway. The social signal transduction theory of depression holds that social stress and rejection, along with early life stress, raise the risk of depression, and that social threat does so in part by increasing proinflammatory cytokines (Slavich & Irwin, 2014). Its standing is that of a hypothesis: the authors present it in a theory review as biologically plausible, describe inflammation as a possible key driver, and offer it as a source of possible explanatory and treatment insights rather than a demonstrated mechanism (Slavich & Irwin, 2014) (see Stress, Adversity and Onset and The Biology of Major Depressive Disorder).

Country income matters for perinatal depression too. A systematic review with meta-regression of 140 prevalence estimates from 96 studies put the pooled prevalence of perinatal depression at 11.9% (95% CI 11.4-12.5) and found higher prevalence in low- and middle-income countries than in high-income ones, as well as higher figures from symptom scales than from diagnostic instruments (Woody et al., 2017). The authors read this as an apparently greater burden in poorer countries, with three limits: studies from low-income countries were scarce, the meta-regression accounted for only 31.1% of the variance, and data were too sparse to pool incidence (Woody et al., 2017).

Social context also governs whether people seek treatment at all. Across 144 studies with 90,189 participants, mental health-related stigma had a small-to-moderate negative effect on help-seeking; the median association was d = -0.27, shrinking to d = -0.18 once low-quality studies were set aside (Clement et al., 2015). Internalized stigma and stigma attached to treatment itself deterred help-seeking most, whereas perceived stigma and personal endorsement of stigmatizing views had negligible effects; the association was clearer for attitudes and intentions (d = -0.52) than for actual help-seeking behaviour, where it was close to zero (Clement et al., 2015). Stigma ranked fourth among reported barriers, and the association studies were of moderate quality, with possible selection bias, and many subgroup comparisons had too few studies to be conclusive (Clement et al., 2015). Across 24 countries in the World Mental Health surveys, a subsample of 63,678 respondents gave their reasons for never seeking treatment or for stopping it, and low perceived need and attitudinal barriers were the dominant reasons (Andrade et al., 2014). Low perceived need was more frequent in mild and moderate cases than in severe ones, and among those who did recognize a need, attitudinal barriers outweighed structural ones for starting and continuing treatment; the balance reversed in severe cases, where structural barriers predominated (Andrade et al., 2014). Women and younger people were more likely to recognize a need, and the authors concluded that barriers of need and attitude are a major problem, making work on mental health literacy important (Andrade et al., 2014).

Sex Differences

Women have nearly twice the odds of a major depression diagnosis that men have, and the gap opens in adolescence. The strongest estimate comes from two meta-analyses of national samples, built on 65 and 95 articles and together covering 1,716,195 and 1,922,064 people from over 90 different nations (Salk et al., 2017). For diagnoses, the odds ratio favouring women was 1.95 (95% CI [1.88, 2.03]); for symptom levels the difference was smaller, at d = 0.27 ([0.26, 0.29]) (Salk et al., 2017). The two figures measure different things: the first compares the odds of meeting diagnostic criteria, the second compares average symptom scores, so a small mean difference in symptoms can sit alongside a near-doubling of diagnoses.

Age was the strongest moderator of the gap. The diagnostic odds ratio had already reached 2.37 at age 12 and rose to 3.02 for ages 13 to 15, while the symptom difference peaked at d = 0.47 at age 16, after which the gap narrowed in adulthood (Salk et al., 2017). Gender equity behaved unexpectedly: nations with greater gender equity showed a larger sex difference in diagnoses, whereas equity was not significantly related to the difference in symptoms (Salk et al., 2017). The authors read the pattern as a health disparity affecting women, and they added that depression in men still warrants attention (Salk et al., 2017).

Explaining the gap has proved harder than measuring it. A narrative review published as a Series paper put women's lifetime risk of developing depression at about twice that of men, yet judged the evidence for risk factors specific to depression to be limited (Kuehner, 2017). Of the gender-related subtypes it weighed, the review suggested that a developmental subtype may best explain the difference between women and men (Kuehner, 2017).

A separate line of research concerns how people respond to low mood once it begins. Response styles theory holds that ruminative responses, a repetitive focus on one's depressed state, lengthen periods of depression, while distraction shortens them; in the laboratory and field studies the theory paper drew on, people who ruminated stayed depressed longer than those who distracted themselves (Nolen-Hoeksema, 1991). That evidence concerns the duration of depressed mood in general. On its own it does not show that rumination accounts for the female excess in diagnoses, and the theory's wider evidence is taken up under Helplessness, Hopelessness and Rumination.

Reproductive transitions add a burden that falls on women specifically. Perinatal depression is common: a systematic review with meta-regression of 140 usable prevalence estimates from 96 studies put its pooled prevalence at 11.9% (95% CI 11.4-12.5) (Woody et al., 2017). Estimates were higher when depression was measured with symptom scales than with diagnostic instruments, and higher in low- and middle-income than in high-income countries, which the authors took to suggest an apparently greater burden in those settings (Woody et al., 2017). The caveats are substantial. Data from low-income countries were scarce, there were too few data to pool an incidence figure, and the meta-regression explained only 31.1% of the variance between studies (Woody et al., 2017). The peripartum presentation is described in Types of Major Depressive Disorder.

Premenstrual dysphoric disorder (PMDD), a mood condition tied to the premenstrual phase of the cycle, shows how much a prevalence figure depends on the diagnostic method. Across 44 studies with 48 independent samples and 50,659 participants, a random-effects meta-analysis found a point prevalence of 3.2% (1.7%-5.9%) when the diagnosis was confirmed, against 7.7% (5.3%-11.0%) when it was only provisional (Reilly et al., 2024). Restricting confirmed cases to community samples lowered the estimate further, to 1.6% (1.0%-2.5%), with far less heterogeneity (I 2 = 26 %, compared with I 2 = 99 % overall) (Reilly et al., 2024). The authors concluded that provisional diagnosis probably inflates prevalence, and they noted that few of the studies applied the full DSM criteria in community settings (Reilly et al., 2024).

Major Depressive Disorder Across the Lifespan

Major depressive disorder is not confined to any one stage of life, but its risk factors, its presentation and the questions researchers ask about it shift with age. In childhood and adolescence the emphasis falls on onset and prevention; in later life it moves to vascular disease and to the relation between depression and dementia.

Depression among young people appears to be becoming more common, and the increase is most marked in girls and young women (Thapar et al., 2022). The same review treats youth depression as heterogeneous and lists several factors that raise its likelihood: a family history of depression, social stressors, chronic physical illness, membership of a sexual minority, and earlier depressive symptoms, anxiety or irritability (Thapar et al., 2022). On prevention, the review judged that the evidence favours programmes aimed at young people already at raised risk or already showing symptoms (indicated or targeted prevention) over universal programmes delivered to everyone, and that emerging school-based interventions show some promise (Thapar et al., 2022). For treatment the authors recommended a stepwise approach, which makes early intervention part of the clinical pathway rather than an optional extra.

The sex difference in depression also has a developmental timetable. Two meta-analyses of national samples, with samples totalling 1,716,195 and 1,922,064 people across over 90 nations, found that women exceeded men on both measures (diagnoses OR = 1.95; symptoms d = 0.27), and that age was the strongest moderator of the gap (Salk et al., 2017). For diagnoses the gap was OR = 2.37 at age 12 and 3.02 at ages 13 to 15; for symptoms it was d = 0.47 at age 16; and in both cases it was larger in adolescence than in adulthood (Salk et al., 2017). The authors read the pattern as a health disparity, while stressing that depression in men still matters; its wider correlates are taken up under Sex Differences.

In later life, the vascular depression hypothesis proposed that cerebrovascular disease may contribute to some depressive syndromes of old age, with prefrontal systems disruption and a threshold of lesion accumulation among its proposed mechanisms (Alexopoulos et al., 1997). Its authors presented the mechanisms as hypotheses, restricted the claim to some geriatric depressive syndromes, cited the co-occurrence of depression with vascular disease and its risk factors as support, and expected the concept to yield clinically useful studies (Alexopoulos et al., 1997).

The clearest later-life finding concerns dementia. Across 23 cohort studies with 49,612 older participants, depression carried pooled risks of 1.85 for all-cause dementia (confidence interval 1.67 to 2.04), 1.65 for Alzheimer's disease (1.42 to 1.92) and 2.52 for vascular dementia (1.77 to 3.59), and in a subgroup analysis of five studies the vascular figure was significantly larger than the Alzheimer's one (P = 0.03) (Diniz et al., 2013). The evidence has limits: the vascular estimate rested on only five studies, and the authors acknowledged that none of the cohorts used structured diagnostic interviews, follow-up ranged from 1 to 17 years, some study samples were not representative or not randomly selected, and hazard and odds ratios were pooled together (Diniz et al., 2013). Adjusted estimates were smaller but remained significant, and the authors framed the result as an association, suggesting that prevention trials may be worthwhile (Diniz et al., 2013). The cognitive side of this link is discussed under Cognitive Impairment and Dementia in Later Life.

One finding holds across these stages. In a meta-analysis of 16 longitudinal cohort studies (n = 67 318), people with subthreshold depression developed major depression at almost twice the rate of non-depressed people (IRR = 1.95, 95% CI 1.28–2.97) (Lee et al., 2019). The ratio did not differ significantly between youth, adults and older people, or between community and primary care samples, which the authors took as support for indicated prevention at any age and in any setting (Lee et al., 2019).

Comorbidity

Major depressive disorder seldom arrives alone, and its company changes how much harm it does. Comorbidity here means the co-occurrence of depression with a second condition in the same person, whether a chronic physical illness or another mental disorder. The two pairings with the most evidence are physical disease and anxiety.

The clearest population picture comes from the World Health Survey, a cross-sectional survey of 245 404 adults in 60 countries that classified depression by ICD-10 criteria and compared mean health scores with regression modelling (Moussavi et al., 2007). Of the conditions examined, depression lowered health scores more than angina, arthritis, asthma or diabetes did. Comorbidity compounded the burden: respondents with depression on top of a chronic physical disease reported worse health than those with depression alone or with the physical disease alone, and depression was more common among people with a chronic disease than among those without one. The authors read the pattern as making depression an urgent public-health priority. Because the survey was cross-sectional, it describes where the burden falls rather than which condition leads to the other.

Stroke offers a specific case. A systematic review and meta-analysis of 61 studies, covering 25 488 people, put the pooled frequency of depression after stroke at 31% (95% confidence interval 28% to 35%) (Hackett & Pickles, 2014). The figure was lower later on, at 25% between one and five years after the stroke and 23% at five years, and estimates varied considerably from study to study. The updated pooled frequency did not differ from the authors' 2005 review, which led them to conclude that post-stroke depression has not yet been reduced.

Among mental disorders, the overlap is broad. NESARC-III, a national cross-sectional survey of 36 309 noninstitutionalized US adults, found major depressive disorder associated with both anxiety and substance use disorders (Hasin et al., 2018). The largest adjusted odds ratio was for generalized anxiety disorder (aOR 5.7; 95% CI, 4.98-6.50); any drug use disorder (aOR 3.0; 95% CI, 2.57-3.55), specific phobia (aOR 2.1; 95% CI, 1.84-2.35) and alcohol use disorder (aOR 1.8; 95% CI, 1.63-2.01) were also linked to it. As a single-time survey, it shows which disorders travel together, not which comes first.

Anxiety is the psychiatric comorbidity with the plainest consequences for treatment. In a secondary analysis of the STAR\*D trial, Fava et al. classified 2,876 adult outpatients with MDD as anxious or nonanxious using an anxiety/somatization factor score drawn from the 17-item Hamilton Depression Rating Scale (Fava et al., 2008). Patients with anxious depression remitted less often and more slowly, reported more side effects, and fared worse when they moved to second-step treatments, so higher baseline anxiety predicted a lower chance of remission. The authors concluded that anxious depression carries poorer acute outcomes, although the comparison came from a secondary analysis.

Twin data suggest that depression and generalized anxiety share much of their inherited basis. Drawing on 37296 twins from the Swedish Twin Registry, with lifetime diagnoses made from interviews by modified DSM-IV criteria, bivariate twin models estimated the genetic correlation between major depression and generalized anxiety disorder at +1·00 in women and +0·74 in men (Kendler et al., 2007). In women, then, the genetic risks were indistinguishable. Correlations for individual-specific environment ran the opposite way by sex, +0·59 in men against +0·36 in women. A trivariate Cholesky model, fitted only to same-sex pairs, added a self-report neuroticism questionnaire: genes for neuroticism raised the risk of both disorders, to a similar degree for each in both sexes, and accounted for around 25% of the genetic covariance the two disorders share. The authors concluded that the genetic overlap is strong and greater in women.

Cognitive vulnerability research suggests the overlap is not symmetrical. In a prospective cohort, 347 college freshmen with no disorder at the outset were grouped as high or low in cognitive risk by their negative cognitive styles and followed for 2.5 years (Alloy et al., 2006). The high-risk group had 3.5 to 6.8 times greater odds of major, minor and hopelessness depression, and the styles predicted recurrences of major and hopelessness depression as well as they predicted first onsets; for minor depression they predicted first onsets more strongly. Anxiety behaved differently: negative cognitive styles did not predict anxiety that occurred without depression, and predicted anxiety onset only when it was comorbid with depression. On this evidence the styles look specific to depression rather than a general risk for distress, though the sample was a single cohort of students. Beck's account of such styles is set out under Beck's Cognitive Model of Depression, and hopelessness under Helplessness, Hopelessness and Rumination.

Depression in the Family

Major depressive disorder is usually described as something that happens to one person, yet it unfolds inside households, and the people who live with a depressed patient have become a research subject in their own right. Two bodies of work are relevant here: quantitative studies of the children of depressed mothers, and qualitative studies of what relatives report about living alongside the illness. Material on depression in young people and on the perinatal period belongs to the sections Major Depressive Disorder Across the Lifespan and Sex Differences, and is not repeated here.

A quantitative synthesis pooled 193 studies of mother-child dyads (Goodman et al., 2011). Children whose mothers were depressed showed higher levels of internalizing problems, externalizing problems and general psychopathology, more negative affect and behaviour, and less positive affect and behaviour. Every one of these associations was small. Their strength was not fixed: moderating variables altered how closely maternal depression and child outcomes were linked, and the pattern of moderation differed from one outcome to another. The authors therefore argued that the field needs explanatory models that go beyond main effects (Goodman et al., 2011). Two limits follow directly from the design. The findings are correlational, so they show that maternal depression and child difficulties co-occur, not that one produces the other or by what route. And the synthesis concerned mothers only, so it says nothing directly about depressed fathers or other caregivers.

The second literature asks what depression is like for the relatives themselves. A scoping review located 34 qualitative publications on the relatives of people with depression, a body of research it described as dominated by a focus on carer burden (Buus et al., 2024). Across the studies the reviewers identified four themes. Relatives described the work of coming to see the illness for what it was, and the feelings it provoked in them. They described how their relationships with the depressed person were broken off or disrupted, and they portrayed their own adjustment as a psychosocial process that moved through stages (Buus et al., 2024). The review was also critical of the research it summarised. Many analyses were de-contextualized and underinterpreted, studies tended to assume that relatives share a single kind of experience, and the social conditions that contribute to that experience were largely disregarded (Buus et al., 2024).

Taken together, the two literatures point in the same direction. Both find that depression reaches beyond the patient, and both report that the existing evidence is narrower than the question it addresses, whether because main-effect models leave out moderators or because a focus on burden leaves out context.

Cognitive Impairment and Dementia in Later Life

Depression that appears or persists in older age is linked with a raised risk of later dementia. A meta-analysis pooled 23 cohort studies covering 49,612 older adults and found that those with late-life depression went on to develop all-cause dementia more often than those without it, with a pooled risk estimate of 1.85 (95% CI 1.67-2.04) (Diniz et al., 2013). The excess held for Alzheimer's disease, at 1.65 (95% CI 1.42-1.92), and was larger for vascular dementia, at 2.52 (95% CI 1.77-3.59); the gap between the two dementia types reached significance (P = 0.03) (Diniz et al., 2013). Studies that used the Center for Epidemiologic Studies Depression Scale gave estimates no different from those using other depression scales (Diniz et al., 2013).

The authors framed the result as an association and suggested that prevention trials in this group may be worthwhile, and several limits temper the figures (Diniz et al., 2013). The vascular dementia estimate rested on only five studies; none of the included cohorts diagnosed depression with a structured interview; some study samples were not representative or not randomly drawn; follow-up ran anywhere from 1 to 17 years; hazard ratios and odds ratios were combined in a single pool; and recurrent episodes were not examined on their own (Diniz et al., 2013). Adjusting for other factors shrank the estimates, although they stayed significant (Diniz et al., 2013). None of this establishes that depression causes dementia: the evidence is a consistent co-occurrence over time, open to more than one causal reading.

Risk Factor or Early Sign?

When symptoms were measured matters for that reading. The Whitehall II cohort tracked 10 189 civil servants, 6838 men and 3351 women, over 28 years, with depressive symptoms assessed on 9 occasions (Singh-Manoux et al., 2017). Symptoms recorded in midlife, in 1985, did not significantly predict later dementia (HR, 1.21; 95% CI, 0.95-1.54), whereas symptoms recorded in 2003, in later life, did (HR, 1.72; 95% CI, 1.21-2.44) (Singh-Manoux et al., 2017). The same split appeared for symptoms that kept returning: a chronic or recurring pattern early in follow-up carried no significant excess risk (HR, 1.02; 95% CI, 0.72-1.44), while one confined to the later years did (HR, 1.67; 95% CI, 1.11-2.49) (Singh-Manoux et al., 2017). Looking backwards from diagnosis, people who developed dementia had higher symptom scores than those who stayed free of it from about 11 years beforehand (Singh-Manoux et al., 2017). The authors therefore read late-life symptoms as a possible prodromal feature of dementia, or as an outcome sharing causes with it, rather than as a risk factor in their own right; dementia was identified from electronic health records, and some participants were lost to follow-up (Singh-Manoux et al., 2017).

The Vascular Route

One proposed link runs through the blood vessels of the brain. Under the vascular depression hypothesis, cerebrovascular disease may contribute to some geriatric depressive syndromes but not to every one, and its proponents took the link between ischaemic lesions and a distinctive set of behavioural symptoms as support (Alexopoulos et al., 1997). Its proposed mechanisms, among them disruption of prefrontal systems and a threshold of accumulated lesions beyond which symptoms appear, were offered as hypotheses, and the authors argued chiefly that the concept could generate useful research (Alexopoulos et al., 1997).

A later narrative review kept the hypothesis in that provisional standing. It set out three candidate pathways from vascular disease to late-life depression, the disconnection, inflammatory and hypoperfusion hypotheses, and treated each as a claim still awaiting a test (Taylor et al., 2013). The review linked executive dysfunction with a weaker response to antidepressants, and linked worsening cerebral hyperintensities and advancing cognitive deficits with a poorer course of depression (Taylor et al., 2013). Its authors suggested that testing these mechanisms might point towards drugs already in use for other conditions that could be repurposed (Taylor et al., 2013). The vascular account therefore remains a working hypothesis about a subgroup of late-life cases rather than a settled explanation of the dementia association.

Do Deficits Outlast the Episode?

Whether cognitive dysfunction clears when mood recovers bears on how the dementia link should be read. A meta-analysis of studies using the Cambridge Neuropsychological Test Automated Battery (CANTAB) found that people with depression scored below controls on executive function, memory and attention, with effects of moderate size (Rock et al., 2014). In remitted patients, executive and attentional deficits were still present, while the memory difference from controls was non-significant (Rock et al., 2014). The authors concluded that cognitive impairment belongs to the core of depression and is not wholly secondary to low mood (Rock et al., 2014). That analysis was not limited to older adults, so it shows that some deficits persist beyond the episode without showing that they progress towards dementia; the broader pattern of attentional, memory and executive findings is set out under Attention, Memory and Executive Biases, and age-related differences under Major Depressive Disorder Across the Lifespan.

Such deficits also bear on everyday functioning, although the evidence here is thin. Across 28 studies of adults with MDD, one systematic review found that weaker executive function, attention, memory and global cognition each went with poorer psychosocial functioning, and its authors suggested that older age and greater severity may strengthen the tie (Cambridge et al., 2018). Only some cognitive domains predicted long-term outcome in that review, and the part played by psychosis or by length of illness was unclear (Cambridge et al., 2018). An earlier narrative review, restricted to 10 studies of nongeriatric adults under 60, saw a broad association between neurocognitive deficits and functional impairment but rated the quality of the evidence as weak, and the heterogeneity of the studies ruled out a meta-analysis (Evans et al., 2014).

Suicide Risk

Assessing the risk of suicide is part of the clinical care of depression, and the question for research is which characteristics of a depressed person raise that risk. A systematic review with meta-analyses addressed it within depressed samples, drawing on nineteen studies reported across 28 publications (Hawton et al., 2013). Seven factors were associated with later suicide. Two concern the person's history: an earlier suicide attempt and psychiatric illness in the family. Three describe the current illness: greater severity of depression, hopelessness and co-occurring anxiety. The remaining two are male gender and the misuse of alcohol or drugs (Hawton et al., 2013). The odds ratios reported for these associations ranged from 1.41 to 4.84, and the authors concluded that the factors should inform clinical risk assessment (Hawton et al., 2013).

The review's limits bear on how the list is read. Fewer eligible studies existed than the authors had expected, and they called for larger ones; nor did they examine how the factors depend on one another (Hawton et al., 2013). Each factor therefore marks elevated risk on its own terms, and the review cannot say how several factors combine in one person.

A risk factor that is reliably associated with suicide is not the same as a tool that can tell who will act. Pooling 3,428 risk factor effect sizes from 365 studies published over 50 years, a random-effects meta-analysis found that the broad categories of risk factors forecast suicidal thoughts and behaviours only slightly better than chance (Franklin et al., 2017). Accuracy had not risen in more recent studies, and studies with longer follow-up predicted no better than short ones (Franklin et al., 2017). The authors traced part of the problem to the research itself, which was methodologically narrow and homogeneous, rarely combined several risk factors and left theories untested, and they argued for a move toward machine-learning risk algorithms (Franklin et al., 2017).

A second meta-analysis narrowed the question to depression and hopelessness, combining 357 prediction cases from 166 longitudinal papers published between 1971 and 2014 (Ribeiro et al., 2018). Both constructs predicted later suicidal ideation, attempts and death, but the associations were weak to moderate. For hopelessness the weighted odds ratios were 2.19 for ideation, 1.95 for attempts and 1.98 for death; for a diagnosis of MDD they were 2.48, 2.38 and 1.50 (Ribeiro et al., 2018). Depressive symptoms predicted ideation and attempts but not death (Ribeiro et al., 2018). Studies differed widely, publication bias was evident and the estimates shrank once it was adjusted for; the authors judged the methodological limits prominent and considered remedying them likely to be worthwhile (Ribeiro et al., 2018).

Hopelessness, one of the seven, is also the organising construct of a cognitive account of depression. The hopelessness theory, set out as a cognitive diathesis-stress model, proposed a subtype it called hopelessness depression, with an attributional diathesis as one specific vulnerability and mediational processes linking stress to symptoms (Abramson et al., 1989). At the time only one part had been tested, the interaction of the attributional diathesis with stress in producing depression; a diathesis for inferring negative consequences, and the claim that cognitive style interacts with stress to produce clinical depression, had not (Abramson et al., 1989). The authors listed further open questions: the symptoms of hopelessness depression had received little study, the stability of the diathesis was unresolved, the theory did not specify a time lag, and hopelessness might prove to be a cause of depression rather than the mark of a subtype (Abramson et al., 1989). Its standing at publication was that of a proposed model, which its authors judged promising while stating that its course, cure and prevention claims still needed testing (Abramson et al., 1989). The theory's wider evidence is taken up under Helplessness, Hopelessness and Rumination.

A later cognitive account puts other constructs at its centre. In the integrated motivational-volitional model, the sense of defeat gives rise to suicidal ideation by way of feelings of entrapment, and a separate set of volitional moderators governs whether ideation turns into an attempt (O'Connor & Kirtley, 2018). In one study of college students its authors reviewed, people who had made an attempt differed from those with ideation alone on volitional-phase variables but not on motivational ones, and in another study entrapment predicted hospital readmission for self-harm (O'Connor & Kirtley, 2018). The authors described the model's support as growing but left its temporal ordering and its interactions unresolved, noting that prospective studies were few, that defeat and entrapment had often been measured at the same moment, and that no evidence-based treatment yet targeted entrapment (O'Connor & Kirtley, 2018).

The interpersonal theory of suicide is the other main ideation-to-action account: on this view, the most dangerous form of suicidal desire arises when thwarted belongingness and perceived burdensomeness are present together, along with hopelessness about these states, whereas the capability for suicide is acquired separately through repeated painful or fearful experience (Van Orden et al., 2010). Its authors put these claims forward as hypotheses to be tested and potentially falsified, and they reflected on the status of theory about suicidal behaviour, indicating areas of relative weakness (Van Orden et al., 2010). When the theory's tests were pooled in a meta-analysis of 122 samples, the predicted interactions held up (thwarted belongingness with perceived burdensomeness for suicidal ideation, and both of these with capability for suicide for prior attempts), although the effects were modest (Chu et al., 2017). The same meta-analysis also found that combinations of the theory's variables other than those it predicts forecast suicide risk about as well (Chu et al., 2017).

Safety planning is one response that has been tested against suicidal behaviour itself. In a nonrandomized cohort comparison spanning nine emergency departments, suicidal behaviour in the six months of follow-up was recorded for 36 of 1186 veterans (3.03%) offered safety planning plus telephone follow-up and for 24 of 454 (5.29%) given usual care, roughly 45% fewer events (OR = 0.56, P = .03), and the intervention group also had roughly twice the odds of making at least one outpatient mental health visit (OR = 2.06, P < .001) (Stanley et al., 2018). The sample was mostly men and the groups were compared as cohorts rather than randomised; the authors concluded that the intervention was associated with a reduction in suicidal behaviour and increased treatment engagement and may be a valuable clinical tool (Stanley et al., 2018). A meta-analysis of six studies with 3536 adults at risk of suicide then pooled safety planning-type interventions: suicidal behaviour was less likely than under control conditions (RR = 0.570, 95% CI 0.408–0.795), a 43% reduction with a number needed to treat of 16, while the difference in suicidal ideation was not significant (g = 0.69, 95% CI -0.04 to 1.42, P = 0.06) (Nuij et al., 2021). That evidence base is thin. The controlled studies were few, not all were randomised, methodological quality was low, follow-up periods varied, no adolescents were included and publication bias was suggested; adjusting for it reduced the estimate to an RR of 0.71 (Nuij et al., 2021). The authors took the findings as support for these interventions and for including them in clinical guidelines for suicide prevention (Nuij et al., 2021).

Anyone who is thinking about suicide can contact a local crisis line or emergency services for immediate help.

The line between the disorder and an understandable reaction to misfortune has been argued most sharply over loss. The question took the form of the bereavement exclusion, a diagnostic rule that treated depression following a death differently from depression following any other loss. Wakefield and colleagues tested the rule's premise in National Comorbidity Survey (NCS) data on 8098 persons aged 15 to 54 years (Wakefield et al., 2007). With the survey's uncomplicated bereavement algorithm, they set uncomplicated depression that followed a death beside uncomplicated depression that followed other losses and compared the two on nine indicators of disorder. On eight of the nine the groups did not differ significantly, and the symptoms each trigger produced looked much alike (Wakefield et al., 2007). The uncomplicated cases of either kind, by contrast, showed fewer signs of disorder than complicated cases on most indicators, and the authors concluded that the NCS gave no support for exempting bereavement alone (Wakefield et al., 2007). In these data, then, the distinction that separated cases was whether the depression was complicated, not which kind of loss had set it off.

A narrative review that added studies published after 2006 reached a related verdict by a different route. Depressive episodes arising after bereavement resembled other major depressive episodes in their genetic influences, their comorbidity, their chronicity and recurrence, and their response to medication; a person's own past episodes and a family history of major depressive episodes both raised the likelihood of an episode after bereavement (Zisook et al., 2012). On that basis the reviewers recommended that DSM-5 drop the exclusion (Zisook et al., 2012). The two sources agree that a death does not by itself mark a depressive episode as benign, though one rests on a single community survey and the other on a narrative synthesis rather than a pooled analysis.

At the chronic end of the spectrum lies persistent depressive disorder. A narrative review approached it dimensionally, organising chronic depression by its course and covering both dysthymia and chronic major depression (Schramm et al., 2020). In the review's account, persistent depressive disorder is a common, chronic condition that is often more disabling than the episodic form of major depression; the authors added that dysthymia tends to go undetected and that little is yet known about what causes chronic depression (Schramm et al., 2020). Its authors judged the causes likely to be multifactorial and proposed a modular approach to treatment (Schramm et al., 2020). The course-based presentations are taken up individually under Types of Major Depressive Disorder.

The border with bipolar disorder is drawn partly through mixed depression and the DSM-5 mixed features specifier. A review combining historical and systematic methods found the specifier's criteria poorly grounded, describing their evidential basis as weak and suggesting they may prove harmful for treatment (Koukopoulos & Sani, 2014). That judgment comes from one review group, and it leaves the boundary itself contested rather than redrawn.

Below the diagnostic threshold sits subthreshold depression, depressive symptoms below the diagnostic threshold. A meta-analysis of 16 longitudinal cohort studies, together including 67 318 participants, found that people with subthreshold depression developed major depression at almost twice the rate of non-depressed people (incidence rate ratio 1.95, 95% CI 1.28–2.97; Lee et al., 2019). The authors estimated similar ratios for youth, adults and older people, and for community and primary care samples (Lee et al., 2019). The authors read the result as favouring indicated prevention, targeted at people with subthreshold symptoms, across ages and settings (Lee et al., 2019); the wide confidence interval leaves the size of the added risk imprecise, although its direction is consistent. Approaches to forestalling later episodes are discussed under Preventing Relapse and Recurrence.

Treatment Evidence

Every antidepressant tested in a large comparative synthesis beat placebo, but how much each drug helps, and for whom, remains disputed. A systematic review and network meta-analysis drew on 522 trials with 116 477 participants and found that all 21 antidepressants it covered produced higher response rates than placebo (Cipriani et al., 2018). The advantage varied by drug: amitriptyline (OR 2·13, 95% CrI 1·89-2·41) separated from placebo further than reboxetine (OR 1·37, 1·16-1·63). Acceptability, measured as dropout, told a partly different story, with fewer dropouts than placebo for agomelatine and fluoxetine and more for clomipramine. In head-to-head comparisons seven drugs, among them agomelatine, escitalopram and mirtazapine, were more effective than the others, while six, among them escitalopram, sertraline and vortioxetine, were better tolerated (Cipriani et al., 2018). These rankings carry real uncertainty. Most comparisons had wide credible intervals, the head-to-head results were more variable, differences between drugs shrank in placebo-controlled trials, and certainty rated with the GRADE framework ran from moderate to very low; only 18% of the trials were judged at low risk of bias. The authors nonetheless held that the findings should guide evidence-based practice and clinical guidelines.

Whether that benefit depends on how ill a patient is at the start is a live disagreement. Kirsch and colleagues' earlier meta-analysis reported that the gap between drug and placebo widened as baseline severity increased, yet it met the usual benchmark for clinical significance only among the most severely depressed patients (Kirsch et al., 2008). The difference widened as baseline severity rose, but the authors traced the widening to weaker placebo response in the most severely ill, not to a stronger response to the drug. In moderate depression drug and placebo did not differ, and the authors judged the gap small even in severe cases. A patient-level meta-analysis of 6 studies, covering 718 adult outpatients, found that the drug advantage over placebo grew with baseline severity: it was substantial in very severe depression, but below a Hamilton Depression Rating Scale score of 23 the difference was small (d < 0.20) (Fournier et al., 2010). Its authors concluded that for mild or moderate depression the benefit may be minimal or nonexistent, on average. A later individual participant data analysis of Japanese placebo-controlled trials (n = 2464) reached the opposite reading (Furukawa et al., 2018). The severity-by-treatment interaction was not significant (coefficient = −0.04, 95% confidence interval: −0.16 to 0.08, P = 0.49), so the authors judged that patients with mild, moderate and severe depression likely gain a similar benefit. Neither study settles the matter: the first rests on a modest pooled sample, and the second obtained data from only six of eleven eligible trials, all from one country.

The largest individual-level dataset recasts the question. Drawing on 232 placebo-controlled trials submitted to the U.S. Food and Drug Administration between 1979 and 2016, with 73 388 participants, one analysis put the average drug advantage at 1.75 points on HAMD17-equivalent scores (95% CI 1.63 to 1.86), a standardised difference of 0.24 in adults and 0.13 in pediatric trials (Stone et al., 2022). Higher baseline severity enlarged that gap, but only slightly, by 0.09 points for each point of severity. When finite mixture models split the outcomes into response classes, a large response appeared in 24.5% of participants on a drug against 9.6% of those on placebo, while minimal response was less common on a drug (12.2% against 21.5%). From these classes the authors suggested that about 15% of patients receive a substantial benefit specific to the drug. The estimate concerns acute treatment only, rests on baseline and end-of-treatment scores, and cannot fully exclude functional unblinding.

The published record also makes antidepressants look better than they are. Comparing the U.S. Food and Drug Administration's reviews of 74 registered trials of 12 antidepressants, involving 12,564 patients, with what appeared in journals, one analysis found that trials with positive outcomes were more likely to be published (Turner et al., 2008). In the journals, 94% of trials looked positive; by the FDA's own analysis, 51% were. Published effect sizes exceeded the FDA estimates by 32% overall. The comparison could not establish whether authors, sponsors or editors were responsible, and the authors warned that selective reporting of this kind may harm both researchers and patients. A reanalysis of the 522 trials behind the 2018 network meta-analysis took a more sceptical view of that evidence base (Munkholm et al., 2019). It estimated the drug-placebo difference at 1.97 points on the 17-item Hamilton depression rating scale (95% CI 1.74 to 2.21; SMD 0.29), and the effect varied with how trials were run and reported: published trials showed larger effects than unpublished ones (SMD 0.33 against 0.15), trials with a placebo run-in showed marginally larger effects (p=0.05), and industry-sponsored trials showed smaller effects than unsponsored ones (0.27 against 0.41). Judging the trials biased and the evidence uncertain, the authors concluded that no definitive conclusion about benefit could be drawn and that efficacy remains unclear. Their reanalysis used the data the original review reported and did not search the literature afresh.

Benefit is only part of the record. In an online questionnaire answered by 1829 adults who had been prescribed antidepressants, more than half reported 8 of the 20 adverse effects asked about (Read et al., 2014). Sexual difficulties (62%) and feeling emotionally numb (60%) were the most frequent, withdrawal effects were reported by 55%, and 39% reported suicidality. Adverse-effect scores were higher among younger, less educated and lower-income respondents and differed by antidepressant type, but were not related to how depressed respondents had been before treatment. Because the data are self-reports from a survey rather than a trial, they describe what long-term users say they experienced; the authors proposed that such effects may be more common than earlier reports suggested.

Psychotherapy faces the same problem, and correcting for it shrinks the effect. A meta-analysis of 369 comparisons from adult depression trials found that the pooled benefit, larger before adjustment, became small once waiting-list control groups were excluded and risk of bias and publication bias were taken into account with the Cochrane risk of bias tool and the trim and fill procedure (Cuijpers et al., 2019). Effects were larger against waiting-list controls than against care as usual, larger in trials at high risk of bias and larger in non-Western countries, whereas neither the number of sessions nor the type of therapy predicted the effect; for psychodynamic therapy the adjusted estimate did not differ from zero. The authors concluded that Eysenck's challenge to psychotherapy's efficacy was probably wrong, while stopping short of certainty, and they noted that not every bias could be adjusted for, that low-risk studies were too few and that heterogeneity was high. A direct test of the publication gap came from trials funded by the U.S. National Institutes of Health between 1972 and 2008 (Driessen et al., 2015). Following 55 funded grants, it found a mean effect of g = 0.52 in published trials (k = 20) against g = 0.20 in unpublished ones (k = 6), so adding the missing data lowered the pooled estimate to g = 0.39. Psychotherapy remained efficacious on that evidence, but its benefit has been overstated in the literature, and because outcome reporting bias was not examined quantitatively, the true effect may be lower still.

Interpersonal psychotherapy, one specific approach, has its own body of trials. Thirty-eight studies with 4,356 patients met the review's criteria, and in the 16 that compared it with a control group it outperformed control conditions with an effect of d = 0.63 (95% CI=0.36 to 0.90), a number needed to treat of 2.91 (Cuijpers et al., 2011). It did not differ significantly from other psychotherapies, whereas pharmacotherapy did somewhat better than interpersonal psychotherapy (d=-0.19) once one outlying study was removed. Adding medication did not improve on interpersonal psychotherapy alone, though the studies of that combination were few, and in preventing relapse the combination did better than medication alone (odds ratio=0.37).

Combining the two approaches does better than either alone. A network meta-analysis of randomized trials, with 101 studies and 11,910 patients, defined response as a 50% improvement in symptoms (Cuijpers et al., 2020). Combined treatment produced more responders than psychotherapy alone (RR=1.27, 95% CI: 1.14-1.39) and more than pharmacotherapy alone (RR=1.25, 95% CI: 1.14-1.37), while psychotherapy and pharmacotherapy did not differ from each other (RR=0.99, 95% CI: 0.92-1.08). On acceptability, both combined treatment (RR=1.23) and psychotherapy (RR=1.17) did better than medication alone. Remission followed the same pattern, and so did the results for chronic and treatment-resistant depression. The authors concluded that combined treatment seems the best choice for moderate depression, and they left long-term effects, cost-effectiveness and specific populations as open questions.

When a first treatment fails, the next steps yield less. Up to four treatment steps, given in sequence, were offered to 3,671 adult outpatients with nonpsychotic major depressive disorder in the STAR*D trial, which counted remission as a QIDS-SR(16) score of 5 or lower (Rush et al., 2006). Remission rates fell from 36.8% at the first step and 30.6% at the second to 13.7% and 13.0% at the third and fourth, for a reported cumulative rate of 67%. Over 12 months of follow-up, patients who had needed more steps relapsed more often, and those who entered follow-up in remission relapsed less than those who did not, although that advantage of remission was no longer present after the third step. A reanalysis of the patient-level data, following the trial's original protocol, put cumulative remission at roughly half the reported figure (Pigott et al., 2023). Analysing 3110 of the 4041 adults enrolled and scoring remission on the Hamilton Rating Scale for Depression, it found a cumulative rate of 35.0% against the reported 67%, with remission at the first level (25.5%) well below the 48.4% seen in comparator trials. The reanalysis is descriptive only, the trial was open label without a control group, and 1330 patients lacked an exit Hamilton score.

Adding an atypical antipsychotic is one option for depression that has not responded. A meta-analysis of 14 short-term placebo-controlled trials (4 to 12 wk) in adults with treatment-resistant major depressive disorder found that all four drugs examined raised remission rates over placebo, and that aripiprazole, quetiapine and risperidone also raised response rates (Spielmans et al., 2013). On clinician-rated severity the advantage was small to moderate (Hedges' g 0.26 to 0.48, about 2.69 points on the Montgomery-Asberg Depression Rating Scale), but quality of life and functioning did not improve beyond placebo, risperidone aside. The drugs also brought harms, including more weight gain than placebo. The authors judged the efficacy modest and urged caution: all but one trial was funded by a manufacturer, and benefits may have been inflated and adverse events underreported.

The evidence for children and adolescents is weaker. A network meta-analysis of 34 trials with 5260 young people with major depressive disorder compared 14 antidepressant treatments and found that only fluoxetine outperformed placebo on efficacy (SMD -0·51, 95% CrI -0·99 to -0·03) (Cipriani et al., 2016). Fluoxetine was also better tolerated than duloxetine and imipramine, whereas imipramine, venlafaxine and duloxetine led to more discontinuations for adverse events than placebo. For most comparisons the quality of evidence was very low, and the authors concluded that antidepressants offer no clear advantage in this age group, with fluoxetine probably the best option to consider. Psychotherapy helps young people only modestly. A meta-analysis of 55 randomized trials with 4,560 participants aged 4 to 18 found an overall effect of g = 0.36, little changed from an earlier estimate of g = 0.34 in 2006, and no significant difference between studies through 2004 and studies after 2004 (Eckshtain et al., 2020). Effects were larger when youths rated themselves than when parents did, larger against waitlists than against placebo controls, and smaller with blinded assessors (0.09 against 0.34), and only half the trials included a follow-up. The authors concluded that the benefit is modest and that stronger treatments are needed.

Behavioral activation and cognitive therapy both belong to the cognitive behavioral therapy family, and one trial suggests they do not perform alike in severe depression. In a randomized placebo-controlled trial of 241 adults with major depressive disorder, initial severity moderated outcome (Dimidjian et al., 2006). Among the more severely depressed patients, behavioral activation matched antidepressant medication, and both outperformed cognitive therapy. The authors discussed what the result means for treatment guidelines and for disseminating therapy. How these treatments are thought to produce change, and how far the mediator evidence supports those accounts, is taken up under How Psychological Treatments Work.

How Psychological Treatments Work

Knowing that a treatment helps is not the same as knowing why it helps. The outcome trials summarised under Treatment Evidence establish the first point for several therapies. The second rests on mediators: changes produced by the treatment that in turn produce the fall in symptoms. The cognitive account of how cognitive therapy works is set out under Beck's Cognitive Model of Depression. A mechanism claim of that kind needs more than an association between thought change and improvement. It needs evidence that the change comes first and carries the effect.

The mediator literature has not yet supplied that evidence in convincing form. A systematic review of 35 studies of psychological mediators in psychotherapy for depression judged the research heterogeneous and its methods unsatisfactory (Lemmens et al., 2016). The authors singled out the causal step, showing that a change in the mediator produces the change in symptoms, as the hardest part of the case to make. They also cautioned that mechanisms of change are likely to be complex and to operate at several levels at once, so that simple single-pathway causal models will not be enough (Lemmens et al., 2016). The gap is an old one. An early appraisal of the cognitive theory of depression found many of its descriptive claims, such as negative self-cognitions and mood-congruent recall, substantiated, while the causal and diathesis-stress claims lacked strong support and had been tested in fewer studies (Haaga et al., 1991). The same review rated the evidence that depressed thinking is specifically inaccurate or illogical as weak (Haaga et al., 1991).

Component studies, which compare a full treatment with versions that leave parts out, press the point further. Reviewing them critically, Longmore and Worrell found that cognitive behaviour therapy gained little when its cognitive techniques were included compared with when they were omitted, and they judged the support for cognitive change as the cause of symptom improvement to be weak (Longmore & Worrell, 2007). Their review examined three empirical anomalies that Hayes had raised for the cognitive account and noted that evidence on early rapid response to treatment was lacking (Longmore & Worrell, 2007).

A cognitive route may also operate in pharmacological treatment. This evidence concerns antidepressant drugs, not psychotherapy, but it bears on the cognitive account. A narrative review of behavioural and functional magnetic resonance imaging studies reported that antidepressants reduce negative emotional bias earlier than they lift mood (Harmer et al., 2009). On this cognitive neuropsychological hypothesis, the drugs first alter emotional processing in limbic and prefrontal circuitry, and mood improves later. The authors described the account as possibly consistent with cognitive theories of depression (Harmer et al., 2009). It does not show that bias change mediates recovery in any given patient.

Targeting the biases directly has so far produced modest returns. Pooling 49 randomized trials of cognitive bias modification for anxiety and depression, one meta-analysis found a small advantage over control conditions across all samples, but in patient samples the effects were mostly not significant (Cristea et al., 2015). The advantage also lost significance once outlying trials were removed and again once the estimate was adjusted for publication bias. The trials tended to be small and of low quality, positive results leaned on a few outliers, and the authors concluded that the true effect could be small or could be absent altogether (Cristea et al., 2015).

Behavioural activation puts a sharper question: is changing thoughts necessary at all? The therapy targets activity and engagement rather than the content of thinking. In a randomized placebo-controlled trial of 241 adults with major depressive disorder, initial severity moderated outcome (Dimidjian et al., 2006). Among the more severely depressed participants, behavioural activation performed about as well as antidepressant medication, and both outperformed cognitive therapy (Dimidjian et al., 2006). The authors discussed the implications for treatment guidelines and for dissemination. A single trial, with the advantage confined to the severe subgroup, does not show that cognitive change plays no part in recovery. It does show that a treatment which does not address thought content directly can match one that does, which complicates any claim that altered cognition is the necessary route.

Behavioural activation's own mechanism is no better established. A systematic review of 14 prospective treatment studies, which between them examined 21 candidate mediators, found only weak evidence that changes in activation or in environmental reward account for the therapy's effect (Janssen et al., 2021). Many of the tests were non-significant, some studies were of poor methodological quality, and the questionnaires used to measure processes and beliefs varied from study to study, so the reviewers drew no firm conclusion about any mediator (Janssen et al., 2021).

Cognitive reactivity offers a more specific target, concerned with what keeps people well rather than with how they recover. The term refers to how far a mild sad mood reactivates dysfunctional attitudes in someone who has recovered. After the first phase of a randomized trial, 99 outpatients in remission from major depressive disorder underwent a sad mood provocation and were followed for 18 months (Segal et al., 2006). Patients who had remitted on medication showed more cognitive reactivity than those who had remitted with cognitive behavioural therapy. Those whose dysfunctional-attitude scores rose by 8 points or more after the provocation relapsed sooner than the rest, regardless of the type of treatment received earlier (Segal et al., 2006). Relapse was tracked with the Hamilton Depression Rating Scale and longitudinal interval follow-up evaluations. The authors concluded that reactivity may be related to vulnerability to relapse (Segal et al., 2006). That makes it a candidate target for the approaches described under Preventing Relapse and Recurrence. The finding is prognostic, though: by itself it does not show that reducing reactivity is what prevents relapse.

Brain Stimulation and Rapid-Acting Treatments

Biological treatments beyond the standard antidepressants are aimed at the limits of those drugs. A narrative review of newer biological approaches to major depressive disorder listed those limits: improvement can take up to 6 weeks to appear, side-effects are common, and some patients need more than one agent (Marwaha et al., 2023). The same review covered the brain and body mechanisms proposed for the newer treatments, though as a narrative review it did not pool their effects. The treatments below are judged mainly by trials of symptom response, and only three of the bodies of evidence here measure cognition directly.

Electroconvulsive therapy has the oldest evidence base. A systematic review and meta-analysis that drew on randomised and observational studies found it effective for depression in the short term (UK ECT Review Group, 2003). Against simulated ECT, real treatment produced a standardised effect size of -0.91 (95% CI -1.27 to -0.54; 6 trials, 256 patients), and against drug treatment the figure was -0.80 (95% CI -1.29 to -0.29; 18 trials, 1144 participants). The authors still hedged that comparison, concluding that ECT is probably more effective than pharmacotherapy. Within ECT, bilateral electrode placement was moderately more effective than unilateral placement (SES -0.32, 95% CI -0.46 to -0.19; 22 trials, 1408 participants), and higher doses beat lower doses. Cognitive function and mortality were among the outcomes the review assessed alongside depressive symptoms.

The cognitive cost of ECT appears to be concentrated in the days right after a course. Pooling 84 studies of 2981 patients tested with standardised cognitive measures, one meta-analysis found that in the first 3 days after treatment, performance fell below the pretreatment level on 72% of the variables examined (Semkovska & McLoughlin, 2010). Between days 4 and 15, most variables were back near their pretreatment values, and beyond 15 days 57% of them sat above baseline. The domains covered included processing speed, working memory, anterograde memory and executive function. Neither electrode placement nor stimulus waveform significantly predicted outcome, and after day 3 no moderator did. The authors read the pattern as recovery to baseline, with improvement in some domains. Retrograde amnesia was the gap: the studies offered no standardised tests of it, and earlier research on the question had been inconsistent.

Ketamine has now been tested head to head against ECT. In an open-label, randomised noninferiority trial at 5 sites, 403 patients with non-psychotic treatment-resistant depression were assigned to ketamine (200) or ECT (203), with ketamine given at 0.5 mg per kilogram (Anand et al., 2023). Response on the Quick Inventory of Depressive Symptomatology-Self-Report reached 55.4% with ketamine and 41.2% with ECT. That 14.2 percentage-point difference (95% CI 3.9 to 24.2) cleared the trial's noninferiority margin of -10 percentage points (P<0.001), so ketamine was judged noninferior to ECT. Memory gave the clearest cognitive contrast. On delayed recall from the Hopkins Verbal Learning Test-Revised, scores changed by -0.9±1.1 in the ketamine group and by -9.7±1.2 in the ECT group, so recall declined far less with ketamine. Quality of life did not differ between the arms. The side-effects did: ECT brought more musculoskeletal effects and ketamine more dissociation. Two features limit the result. Neither patients nor clinicians were blinded, and 38 patients withdrew before receiving any treatment.

The drug also comes in two forms, and they may not be equivalent. A systematic review and meta-analysis of randomised trials (24 trials, 1877 participants with unipolar or bipolar depression) compared racemic ketamine with esketamine (Bahji et al., 2021). The pooled risk ratio for response was 3.01 for racemic ketamine and 1.38 for esketamine, and for remission it was 3.70 against 1.47. Dropout ran the other way: 0.76 for ketamine and 1.37 for esketamine. The authors kept their conclusion tentative, saying that intravenous ketamine possibly outperforms intranasal esketamine. Esketamine on its own has a phase 3 test behind it. At 39 outpatient referral centres, 227 adults with treatment-resistant depression were randomised under double-blind conditions to esketamine (56 or 84 mg) or placebo, each given with an antidepressant (Popova et al., 2019). By day 28 the esketamine group had improved more on the Montgomery-Åsberg Depression Rating Scale, a difference of -4.0 (SE=1.69, 95% CI=-7.31, -0.64), and the authors took this as support for esketamine as a rapid-acting antidepressant in this group. Adverse events were more frequent with esketamine, among them dissociation, nausea, vertigo, dysgeusia and dizziness, though they had resolved by 1.5 hours; 7% of the esketamine group discontinued, against 0.9% of the comparison group. Esketamine's regulatory path has also drawn criticism. Turner set out seven concerns about the evidence for esketamine's efficacy in treatment-resistant depression and about its approval by the FDA (Turner, 2019).

Most forms of repetitive transcranial magnetic stimulation (rTMS) beat sham stimulation in a network meta-analysis of randomised trials with 4233 patients with major depressive disorder (59.1% women, mean age 46 years) (Brunoni et al., 2017). For response, the odds ratios against sham ranged from 2.37 to 4.66 across priming low-frequency, bilateral, high-frequency, θ-burst and low-frequency stimulation. Newer rTMS variants did not differ significantly from sham, and θ-burst stimulation did not significantly improve remission. Every intervention was at least as acceptable as sham, judged by dropout. The authors found few differences between modalities and called for larger trials. They also noted that the estimates were imprecise and that some interventions rested on few trials.

Stopping antidepressants raises a separate question, and two reviews disagree about how big a problem it is. A meta-analysis of 79 studies (44 randomised trials and 35 observational studies; 21 002 patients) found discontinuation symptoms more common after stopping an antidepressant than after stopping placebo (Henssler et al., 2024). The incidence of at least one symptom was 0·31 (95% CI 0·27-0·35) after antidepressants and 0·17 (0·14-0·21) after placebo, and severe symptoms occurred in 0·028 against 0·006. Taking the placebo rate as a measure of non-specific effects, the authors put the incidence of discontinuation symptoms at approximately 15%, affecting about one in six to seven patients who stop their medication. They presented this as the probable extent of discontinuation symptoms, one that need not cause undue alarm. Drug type predicted how often symptoms occurred. Venlafaxine and desvenlafaxine produced them most often, and imipramine and paroxetine produced the most severe ones. Diagnosis and trial characteristics did not account for the incidence. Heterogeneity was substantial, and some reported symptoms may have been residual or returning psychopathology rather than withdrawal.

An earlier systematic review of 24 studies took a much more serious view (Davies & Read, 2019). It found withdrawal common, often severe and often prolonged, with incidence ranging from 27% to 86% across studies. Its authors argued that this incidence contradicts guideline descriptions of withdrawal as self-limiting and that symptoms often outlast the two weeks guidelines suggest. They concluded that guidelines probably lead clinicians to misdiagnose withdrawal and need urgent revision. That review also had limits. Its studies varied in method, sample size and population, and only four reported a mean duration. The two reviews therefore leave the size and duration of withdrawal unsettled. The later one subtracts placebo rates to estimate a net effect, while the earlier one emphasises frequency, severity and duration.

Preventing Relapse and Recurrence

Recovery from an episode does not end the clinical problem, and two strategies dominate the evidence on keeping people well: continuing antidepressant medication and teaching a psychological skill set before the next episode arrives. One test of the first came from a double-blind, placebo-controlled trial in UK primary care. Lewis and colleagues randomised 478 adults from 150 general practices in UK primary care to either stay on their medication (238) or discontinue it (240); the sample was 73% women, with an average age of 54 (Lewis et al., 2021). Over 52 weeks, relapse occurred in 135 of 240 people in the discontinuation arm (56%) against 92 of 238 in the maintenance arm (39%), a hazard ratio of 2.06 (95% CI 1.56 to 2.70, P<0.001) (Lewis et al., 2021). The discontinuation arm also reported more depressive, anxiety and withdrawal symptoms. One feature tempers the reading: adherence to the allocated regimen was lower after discontinuation (52%) than under maintenance (70%), so the two groups did not stay on protocol equally (Lewis et al., 2021).

Mindfulness-based cognitive therapy (MBCT) is the psychological alternative considered here. An individual-patient-data meta-analysis pooled 1258 patients with recurrent depression from 9 studies (mean age 47.1 years, 75.0% female) and followed them for 60 weeks (Kuyken et al., 2016). Those who received MBCT had a lower risk of relapse than those who did not (hazard ratio 0.69, 95% CI 0.58-0.82), and the advantage held, at a smaller size, against active treatments (hazard ratio 0.79, 95% CI 0.64-0.97). None of age, sex, education, relationship status, age at onset or number of prior episodes moderated the benefit; there was some evidence that it was larger for people with more severe residual symptoms at baseline. The authors judged that MBCT appears efficacious for relapse prevention, particularly for those with more pronounced residual symptoms, and made recommendations for addressing remaining uncertainties (Kuyken et al., 2016).

Pooled superiority over no MBCT does not mean superiority over medication. In a single-blind trial across 95 general practices, 424 adults with recurrent depression were allocated 212 apiece to a version of the programme labelled MBCT-TS or to maintenance antidepressants and followed for 24 months (Kuyken et al., 2015). Time to relapse or recurrence did not differ between arms (hazard ratio 0·89, 95% CI 0·67-1·18, p=0·43), nor did serious adverse events, and both groups showed lasting positive outcomes. Participants knew which treatment they had received, and the authors concluded that the trial gave no evidence that MBCT-TS was superior for preventing relapse (Kuyken et al., 2015).

A dismantling trial asked which ingredient of MBCT does the work, and its answer complicates the pooled picture. Williams and colleagues allocated 274 adults in remission from recurrent major depression in a 2:2:1 ratio to MBCT, to an active comparison called cognitive psychological education or to treatment as usual, and analysed the 255 retained to follow-up (99, 103 and 53), with MBCT delivered in eight weekly classes and 12 months of follow-up (Williams et al., 2014). Relapse rates were 46%, 50% and 53% respectively, and MBCT did not significantly outperform either cognitive psychological education (HR = 0.88, 95% CI [0.58, 1.35]) or usual care (HR = 0.69, 95% CI [0.42, 1.12]). Childhood trauma, measured with the Childhood Trauma Questionnaire, changed the picture. Among participants with high trauma scores, relapse ran at 41% with MBCT, 54% with cognitive psychological education and 65% with usual care, and MBCT beat usual care (HR = 0.43, 95% CI [0.22, 0.87]) though not cognitive psychological education; among those with low trauma scores the three arms did not differ (Williams et al., 2014). The median-split subgroup analyses were exploratory and had lower power than the full model, homework was not matched across arms, the sample was mostly Caucasian, and a strong, unstandardised usual-care condition may have masked an effect, so the trauma result is a hypothesis to test rather than a settled rule for allocating treatment (Williams et al., 2014).

Why a cognitive programme should protect against relapse is usually explained by cognitive reactivity: the degree to which a mild sad mood reactivates dysfunctional attitudes in someone who has recovered. Segal and colleagues tested it in 99 remitted outpatients aged 18 to 65 who had been treated for major depressive disorder with either medication or cognitive behavioural therapy (Segal et al., 2006). After a sad mood provocation, those treated with medication showed more cognitive reactivity than those treated with cognitive behavioural therapy. Over 18 months of follow-up, assessed with the Hamilton Depression Rating Scale and longitudinal interval follow-up evaluations, people whose dysfunctional-attitude scores rose by 8 points or more under the induced mood relapsed sooner than the rest, regardless of the type of earlier treatment (Segal et al., 2006). The authors framed the link cautiously: reactivity may relate to vulnerability to relapse (Segal et al., 2006). The account is a proposed mechanism; the trials above tested whether MBCT prevents relapse, not whether it does so by reducing reactivity. How psychological treatments are thought to act more generally is taken up in How Psychological Treatments Work.

Worked Example

The diagnosis is polythetic: it is met by any sufficiently large subset of a symptom list rather than by one required set. Counting those subsets shows how much variety the rule permits. The version used here is deliberately simplified. It keeps only the count, requiring at least five of nine symptoms, and the core requirement that at least one of the five be low mood or lost interest or pleasure. The full criteria also set a duration and require distress or impairment; the MeSH definition summarised in What Major Depressive Disorder Is gives the count and the time window.

Step one counts every set of five or more symptoms drawn from nine, ignoring the core requirement. There are 126 sets of five, 84 of six, 36 of seven, 9 of eight and 1 of nine, for 256 in total. Step two removes the sets that contain neither core symptom, which means drawing five or more from the seven non-core symptoms alone: 21 sets of five, 7 of six and 1 of seven, or 29. Step three subtracts, leaving 256 − 29 = 227 qualifying combinations. Broken down by size, 105 have five symptoms, 77 have six, 35 have seven, 9 have eight and 1 has all nine. The PolytheticDemo carries out the same count interactively.

Figure 1

Qualifying Symptom Combinations by Number of Symptoms

Bar chart of the 227 combinations that satisfy a simplified count rule (at least five of nine symptoms, at least one core symptom): 105 with five symptoms, 77 with six, 35 with seven, 9 with eight and 1 with nine. 0 50 100 105 77 35 9 1 5 6 7 8 9 Number of symptoms present
Arithmetic illustration of a simplified count rule; the bars sum to 227. The counts are combinatorial, not patient data.

The arithmetic implies that two people can both qualify while having only one symptom in common. One profile might combine low mood, fatigue, worthlessness, poor concentration and thoughts of death; another might combine lost interest, a change in appetite, a change in sleep, slowed or agitated movement and fatigue. Fatigue is the only overlap, yet both meet the rule.

Whether this variety is merely theoretical is an empirical question. When more than 1,500 patients with DSM-IV major depressive disorder were assessed with semi-structured interviews, 170 of the 227 possible combinations turned up in the sample, so 57 never appeared (Zimmerman et al., 2015). The authors concluded that heterogeneity among diagnosed patients is large and that only a few combinations behave like prototypes (Zimmerman et al., 2015). Two people who share a diagnosis therefore need not share a symptom picture.

Current Directions

Genetic research on major depressive disorder has moved from a handful of favoured genes to genome-wide samples of hundreds of thousands of people. The earlier strategy did not survive testing. Using preregistered analyses in subsamples of 62,138 to 443,264 people, a large study examined 18 candidate genes that had each been studied 10 or more times. It found no support for the candidate polymorphisms or for gene-level effects, and none for interactions between those polymorphisms and environmental adversity such as childhood abuse or socioeconomic hardship. The candidate genes were no more strongly associated with depression than genes nobody had proposed. The authors concluded that the early hypotheses were wrong and that the original positive reports were probably false positives, and they judged phenotypic measurement error an unlikely explanation for the null results (Border et al., 2019). The genome-wide approach has produced replicable signals instead. A meta-analysis of genome-wide association studies from three cohorts covered 807,553 individuals (246,363 cases and 561,190 controls). The analysis turned up 102 independent genetic variants linked to depression, together with 269 genes and 15 gene-sets, and an enrichment analysis pointed to the importance of prefrontal brain regions (Howard et al., 2019). In a replication sample of 1,306,354 individuals, 87 of the 102 variants held up and 15 did not. Mendelian randomization analyses in the same study suggested a putative, not established, causal path from neuroticism to depression. An apparent effect in the opposite direction weakened once overlapping instrumental variants were removed, and depression in turn showed a putative effect on ever having smoked. The authors presented the work as an advance in mapping the disorder's genetic architecture and a possible route to new treatments through drug repositioning.

Inflammation has become a candidate treatment target. An earlier review reported higher peripheral inflammatory biomarkers in people with major depressive disorder than in controls. It linked high inflammation to treatment resistance and described psychosocial stress as a driver of inflammatory signalling. In that account, inflammatory cytokines act through pathways such as nuclear factor kappa B on neuroendocrine function, neural plasticity and glutamate release and reuptake. The authors suggested that immune-targeted therapy may help some patients but stressed that the data on cytokine inhibition were preliminary (Miller et al., 2009). A later systematic review and meta-analysis estimated how common low-grade inflammation is. Across 37 studies of 13,541 depressed patients and 155,728 controls, 27% (95% CI 21–34%) of patients had C-reactive protein (CRP) above 3 mg/L and 58% (95% CI 47–69%) had CRP above 1 mg/L. The odds of exceeding each threshold were higher than in controls (OR 1.46 and OR 1.47 respectively). Sample source, antidepressant treatment, age, body mass index and ethnicity did not significantly account for the prevalence, although estimates varied widely between studies. The authors concluded that inflammation could be relevant to a substantial share of patients (Osimo et al., 2019). Neither source shows that lowering inflammation relieves depression. That question is the one immune-targeted trials would have to settle.

Rapid-acting treatments are a third front, and they have critics as well as supporters. An open-label randomized noninferiority trial at five sites randomized 403 patients with non-psychotic treatment-resistant depression to ketamine (200) or electroconvulsive therapy (ECT; 203). Response on the Quick Inventory of Depressive Symptomatology-Self-Report was 55.4% with ketamine and 41.2% with ECT, a difference of 14.2 percentage points (95% confidence interval, 3.9 to 24.2). Against a noninferiority margin of -10 percentage points, ketamine met the noninferiority criterion (P<0.001). On the Hopkins Verbal Learning Test-Revised, delayed recall declined less with ketamine. The arms did not differ in quality of life. ECT produced more musculoskeletal adverse effects, and ketamine produced more dissociation. Patients and clinicians knew which treatment was given, and 38 patients withdrew before treatment began (Anand et al., 2023). The approval of esketamine has been contested. Turner set out seven concerns about the efficacy evidence for esketamine in treatment-resistant depression and about the United States Food and Drug Administration's approval of it (Turner, 2019). Brain Stimulation and Rapid-Acting Treatments covers these interventions in more detail.

A fourth direction treats the individual symptom, rather than the total score, as the unit of study. A narrative review argued that depressive symptoms differ from one another, so adding them into a sum-score discards information. Psychomotor retardation, for example, was linked to about four times the impairment of agitation. In the studies reviewed, hopelessness predicted suicide better than a sum-score did, insomnia predicted poorer treatment response, and different types of life event preceded different symptoms. The authors called for symptom-level analysis, the study of causal links between symptoms as in the network approach to psychopathology, and treatment tailored to the person. They acknowledged that the results require replication, that existing scales were not built for symptom-level work, and that sum-scores still carry general information (Fried & Nesse, 2015). Heterogeneity and Symptom Networks develops this approach.

Prevention aimed at people below the diagnostic threshold is a fifth. Subthreshold depression refers to depressive symptoms that fall short of the criteria for a major depressive episode. Indicated prevention targets people who already show such symptoms. A meta-analysis of 16 longitudinal cohort studies (n = 67,318) found that people with subthreshold depression developed major depression at almost twice the rate of non-depressed people (incidence rate ratio 1.95, 95% CI 1.28–2.97). The ratio did not differ significantly between youth, adults and older adults, or between community and primary care samples. The authors read this as support for indicated prevention at any age and in either setting (Lee et al., 2019). Preventing Relapse and Recurrence covers prevention after a first episode.

Discussion

Some conclusions about major depressive disorder are secure. It is common and disabling, severe or very severe in about half of the cases a national survey identified, and frequently left without adequate treatment (Kessler et al., 2003). At the group level, depressed people orient towards negative material and resolve ambiguity negatively, although the attention effect was only marginal on the emotional Stroop and the interpretation effect held only for direct measures (Peckham et al., 2010; Everaert et al., 2017). Deficits in executive function and attention are measurable during an episode and can remain after it (Rock et al., 2014). How far such test deficits reach into daily life is less certain: a systematic narrative review found a broad association between neurocognitive deficits and impaired psychosocial functioning but rated the evidence weak, because the studies were few and too varied to pool (Evans et al., 2014). Treatments help: every antidepressant in a network meta-analysis of 522 trials beat placebo, though with certainty rated moderate to very low (Cipriani et al., 2018).

The causal role of cognition is far less settled. An early appraisal of Beck's theory judged much of its account of depressive thinking to hold up, though evidence that depressed people reason in distorted or faulty ways was weak, and the causal diathesis-stress hypotheses had found no strong backing (Haaga et al., 1991). Later designs narrowed the gap without closing it. In a three-wave population cohort, people who went on to a first episode were already more vulnerable beforehand, and their vulnerability returned to its earlier level after recovery with no sign of a lasting scar, a pattern the authors read as a stable trait running alongside a temporary state, with the caution that residual and prodromal symptoms could bias such comparisons (Ormel et al., 2004). Data of this kind fix the order of events but not the mechanism, and the one experimental approach, training biases directly, has not yet shown that it changes depression. Therapy research is more mixed: in studies of cognitive behavioural therapy that addressed temporal order, cognitive change predicted subsequent symptom change, and the reviewers concluded that cognitive change contributes to symptom change, yet few studies ran formal mediation tests, and the reviewers described methodological and inferential limitations in the evidence (Lorenzo-Luaces et al., 2015). Beck's model therefore stands as supported for description and only partly supported, and still debated, as a causal account.

How large treatment effects really are is a second open question. When journal reports of antidepressant trials were set against FDA reviews of the same registered studies, 94% of the published trials appeared positive against 51% by the FDA's analysis, and published effect sizes ran 32% higher overall (Turner et al., 2008). Psychotherapy shows a parallel gap: among the funded trials, published ones averaged g = 0.52 while unpublished ones reached only a non-significant g = 0.20, and folding in the unpublished data brought the pooled estimate down to g = 0.39, a figure the authors caution may still overstate the real effect (Driessen et al., 2015). Both treatments remain effective on this evidence, but by less than the published literature implies. The biology is contested in a similar way. A preregistered umbrella review found no consistent support for a lowered-serotonin cause (Moncrieff et al., 2022), while a critical commentary argued that tryptophan depletion and imaging evidence still implicate the serotonin system, conceding that several of the supporting studies came from the same group and that confidence intervals were wide (Jauhar et al., 2023). The serotonin hypothesis is therefore contested rather than either confirmed or refuted. The environmental record is steadier at one point and emptier at another. In a co-twin analysis of women, stressful life events predicted onset, an association the authors read as substantially causal, with roughly a third of it noncausal (Kendler et al., 1999), whereas pooling 31 datasets found no strong interaction between the 5-HTTLPR genotype and stress (Culverhouse et al., 2018).

Who benefits from medication is disputed as well. A patient-level meta-analysis of 718 outpatients found the drug advantage small below a Hamilton score of 23, and its authors judged that in mild or moderate depression the benefit may be minimal or nonexistent on average (Fournier et al., 2010). Individual data from six of eleven eligible Japanese trials, by contrast, showed no significant interaction between severity and treatment (Furukawa et al., 2018). A participant-level dataset of 232 placebo-controlled trials submitted to the FDA found that the drug-placebo gap increased significantly with greater baseline severity, and its authors suggested, for acute treatment only, that about 15% of patients receive a substantial benefit specific to the drug (Stone et al., 2022). Sequenced treatment calls for similar caution: a descriptive reanalysis of the STAR*D patient-level data put cumulative remission at 35.0% rather than the 67% originally reported (Pigott et al., 2023).

Keeping people well is where the treatment evidence is most direct. In a double-blind primary care trial, relapse over 52 weeks was more common among people who stopped their maintenance antidepressant than among those who continued it (56% against 39%), although adherence to the allocated regimen was lower in the discontinuation arm (Lewis et al., 2021). Mindfulness-based cognitive therapy lowered the risk of relapse in pooled individual patient data (Kuyken et al., 2016), but a single-blind head-to-head trial gave no evidence that it was superior to maintenance medication (Kuyken et al., 2015), and in a dismantling trial it beat usual care only among participants with high childhood trauma, in an exploratory moderator analysis (Williams et al., 2014). The people receiving these treatments add a further perspective. In a qualitative meta-synthesis of 27 studies, participants often felt that antidepressants were the only option open to them and experienced a biochemical-deficit explanation as stigmatising, and the authors called for choices, information and informed decisions about treatment (Crowe et al., 2023).

Whether the specifiers mark distinct conditions is also unresolved. Melancholia has been argued to be a separate mood disorder, more biologically homogeneous than major depression and more specific in its treatment response (Parker et al., 2010), and a meta-analysis of melancholic and non-melancholic patients found worse attention, visual learning and reasoning in melancholia, which its authors read as a moderate but specific melancholic effect, although sensitivity analyses detected no difference among drug-free patients (Zaninotto et al., 2016). Cognition also failed to separate unipolar from bipolar patients during depressive episodes, and in euthymia only verbal memory favoured the unipolar group (Samamé et al., 2017). Heterogeneity limits every average reported on this page. The criteria admit 170 symptom combinations observed in practice (Zimmerman et al., 2015), and in a cohort of 1809 adults every univariate brain difference between patients and controls was tiny, with partial η2 between 0.004 and 0.017, so that the two groups' distributions overlapped heavily and individual diagnosis from the scans was not possible (Winter et al., 2022). A group difference, however reliable, describes the average patient and need not describe any particular one.

Who has the disorder, and alongside what, qualifies the averages further. Women's odds of a diagnosis are close to twice men's and the gap was greater in adolescence than in adulthood (Salk et al., 2017), yet a Series review judged the evidence for risk factors specific to depression limited (Kuehner, 2017), so the gap is better measured than explained. In a Swedish twin sample, the genetic correlation between major depression and generalized anxiety disorder was +1.00 in women and +0.74 in men (Kendler et al., 2007). The illness also reaches beyond the patient: across 193 studies, children of depressed mothers showed more internalizing and externalizing problems, although every association was small and the design correlational (Goodman et al., 2011). In later life, timing matters for interpretation. In the Whitehall II cohort, depressive symptoms predicted dementia when recorded in later life but not when recorded in midlife, which the investigators read as a possible prodromal feature, or an outcome sharing causes with dementia, rather than a risk factor in its own right (Singh-Manoux et al., 2017).

Newer lines of research show the same pattern of real group differences whose meaning is unsettled. Depressed groups performed worse on behavioural reward tasks by a small-to-medium margin, but every pooled study was case-control, heterogeneity was significant and publication bias was evident (Halahakoon et al., 2020), and one conceptual review held that a refined definition of anhedonia, separating the pursuit of reward from its enjoyment, may be essential for progress (Treadway & Zald, 2011). For suicide, broad categories of risk factors predicted thoughts and behaviours only slightly better than chance across 50 years of studies (Franklin et al., 2017), so a list of associated factors is not a tool for telling who will act. Social position shapes both risk and care: the lowest socioeconomic groups had higher odds of depression than the highest, with the gap widest for persisting depression (Lorant et al., 2003), and mental health-related stigma had a small-to-moderate negative effect on help-seeking (Clement et al., 2015).

Common Misconceptions

Depression is caused by low serotonin.
A systematic umbrella review of 17 reviews and large studies found no consistent evidence that lowered serotonin causes depression (Moncrieff et al., 2022). Critics answered that tryptophan depletion and PET findings still point to a role for the serotonin system (Jauhar et al., 2023). Neither side shows that a simple serotonin deficit explains the disorder; the dispute is over whether serotonin is involved at all.
Everyone with depression has the same symptoms.
The criteria can be met by many different subsets of symptoms. In more than 1,500 patients, 170 of the 227 permitted combinations occurred, and only a few were common enough to count as prototypes (Zimmerman et al., 2015). Two people with the same diagnosis may share few symptoms.
Depressed people see the world more accurately than everyone else.
The depressive realism effect exists but is small overall, and it depends on the paradigm, on whether an objective standard of reality was available and on how symptoms were measured (Moore & Fresco, 2012). The belief rests on a real finding, that nondepressed people show more positive bias, stretched into a claim of general accuracy that the data do not support.
A brain scan can diagnose depression.
Across structural, diffusion, task and resting-state imaging in 1809 adults, patients differed from controls only by tiny effects, and brain measures could not predict an individual's diagnosis (Winter et al., 2022). Imaging differences are real at the level of group averages, which is why they are often reported, but the groups overlap too much for a scan to identify a patient.

Glossary

Anhedonia.
A loss of interest in, or pleasure from, activities that were once enjoyed; a core symptom of the disorder, whose standard definition one review argued folds the pursuit of reward together with its enjoyment.
Attentional bias.
Preferential orienting towards negative material, usually measured with the dot-probe task or the emotional Stroop.
Behavioral activation.
A psychological treatment that targets activity and engagement rather than the content of thought; in one trial it matched antidepressant medication among more severely depressed adults.
Bereavement exclusion.
A former diagnostic rule that treated depression following a death differently from depression following other losses, debated during the DSM-5 revision.
Cognitive behavioral therapy for insomnia (CBT-I).
A psychological treatment for insomnia; in patients whose insomnia accompanied another mental disorder it lowered insomnia severity and depressive symptoms after treatment, though effects on comorbid symptoms were not significant at follow-up.
Cognitive reactivity.
The degree to which a mild sad mood reactivates dysfunctional attitudes in someone who is not currently depressed; in one remitted sample, a larger rise predicted earlier relapse.
Comorbidity.
The co-occurrence of depression with a second condition in the same person, whether a chronic physical illness or another mental disorder such as generalized anxiety disorder.
Default mode network.
A large-scale brain network whose connectivity with the subgenual prefrontal cortex has been linked to rumination, and whose internal connectivity was reduced in recurrent but not first-episode, drug-naive depression in one consortium.
Diathesis-stress.
A class of account in which a pre-existing vulnerability produces disorder only when it meets stress.
Dysfunctional beliefs.
In Beck's cognitive model, enduring negative assumptions about the self that are said to form early and to be activated by stressors matching their content.
Entrapment.
In the integrated motivational-volitional model of suicidal behaviour, the state through which defeat is proposed to lead to suicidal ideation; support for the model is growing, but the temporal order of its factors is unresolved.
Experience sampling methodology.
Repeated assessment of people in their daily lives, offered as a complement to laboratory testing; one review of studies with clinically depressed individuals judged it informative while noting methodological limitations.
Hopelessness depression.
A theory-defined subtype identified by its proposed cause, hopelessness arising from a negative attributional style that meets stress, rather than by its symptoms.
Idiom of distress.
A culturally shared way of expressing suffering, such as thinking too much, which one review found should be treated neither as a stand-in for any one psychiatric disorder nor as a single uniform concept.
Interpretation bias.
A tendency to resolve ambiguous information in a negative direction, or to lack the positive reading that nondepressed people tend to show.
Kindling hypothesis.
The proposal that later episodes depend less on major life stress than first episodes do; one review called it the primary framework for this research but found the evidence inconsistent.
Learned helplessness.
An account that linked depression to experience of noncontingency between responses and outcomes, later reformulated so that a person's explanation for uncontrollable events shapes what they expect next.
Low-grade inflammation.
A modest rise in inflammatory markers, often defined as C-reactive protein above 3 mg/L, found in 27% of depressed patients in one meta-analysis.
Melancholia.
A clinical picture recognised within major depression by a specifier, which some researchers argue should be classified as a separate mood disorder.
Mindfulness-based cognitive therapy.
A group programme for people with recurrent depression, delivered in eight weekly classes in one trial, intended to prevent relapse.
Mixed features.
A DSM-5 specifier for manic or hypomanic symptoms that occur during a depressive episode.
Mood-congruent recall.
The tendency to remember affectively toned material whose tone matches the current mood.
Overgeneral memory.
Autobiographical retrieval that returns summarised categories of events in place of specific episodes.
Polygenic risk score.
A single score summarising a person's genetic liability across many variants; in one imaging cohort it separated patients from controls only weakly, about as well as brain measures did.
Polythetic definition.
A category satisfied by any sufficient subset of its criteria rather than by one required set, so that two members may share few features.
Reward anticipation.
The phase in which a possible reward is awaited, as distinct from the consummatory phase in which it arrives; in one imaging study, depressed adults differed from controls less clearly during anticipation than when rewards were delivered.
Rumination.
Repetitive focus on one's depressed state; response styles theory holds that it lengthens depressed mood, whereas distraction shortens it.
Safety planning.
A family of suicide-prevention interventions; pooled across six studies they reduced suicidal behaviour but not significantly suicidal ideation, in an evidence base of low methodological quality.
Scar hypothesis.
The proposal that an episode leaves lasting damage that raises the risk of later episodes; a Dutch population cohort found no such scar effect.
Somatization.
Distress, most often depression, experienced and presented through bodily complaints; the contrasting pattern of presenting distress in emotional terms is called psychologization.
Specifier.
A label added to the diagnosis to describe features of the current episode or its course, such as melancholic, atypical or seasonal features.
Treatment-resistant depression.
Depression that has not responded to treatment, a research construct with no consensus definition; about half of the definitions in one review required at least two failed treatments.

Key Researchers

Aaron T. Beck

(1921–2021). Psychiatrist at the University of Pennsylvania who introduced an inventory for measuring depression and the cognitive model of depression, which he later revised with Keith Bredemeier into a unified model. Wikipedia

Pim Cuijpers

Clinical psychologist at Vrije Universiteit Amsterdam whose meta-analyses examined psychotherapy for depression, its comparison and combination with medication, and the effect of bias on its apparent size. ORCID - Faculty Page

Ian H. Gotlib

Psychologist at Stanford University who, with Jutta Joormann, reviewed cognition and emotion regulation in depression, and who studied rumination and default-mode network connectivity. ORCID - Wikipedia - Faculty Page

Jutta Joormann

Psychologist at Yale University and coauthor, with Ian Gotlib, of a review of cognition and emotion regulation in depression. ORCID - Faculty Page

Kenneth S. Kendler

Psychiatrist at Virginia Commonwealth University and its Virginia Institute for Psychiatric and Behavioral Genetics who coauthored a meta-analysis of the genetic epidemiology of major depression and compared the DSM criteria with classic descriptions of depression. ORCID - Wikipedia

Ronald C. Kessler

Researcher in the Department of Health Care Policy at Harvard Medical School who led national and cross-national surveys of the prevalence, impairment and treatment of major depression. ORCID - Wikipedia - Faculty Page

Susan Nolen-Hoeksema

(1959–2013). Psychologist at Yale University who proposed the response styles theory of rumination in depression. Wikipedia - Institutional Page

Martin E. P. Seligman

Psychologist at the University of Pennsylvania and coauthor of the attributional reformulation of learned helplessness in humans. Wikipedia - Faculty Page

Frequently Asked Questions

What is the difference between major depressive disorder and depression?

Major depressive disorder is a specific diagnosis: the DSM-5-TR requires a minimum number of symptoms lasting a minimum period, at least one of two core symptoms, and clinically significant distress or impairment (American Psychiatric Association, 2022). Depression in the wider sense also refers to low mood, and to depressive symptoms that occur across many conditions, without these requirements being met.

How is major depressive disorder diagnosed and measured?

A clinician judges whether the diagnostic criteria are met, and questionnaires grade severity. The PHQ-9, a nine-item self-report measure, detected major depression with 88% sensitivity and 88% specificity at a score of 10 or more in primary care and obstetrics-gynecology patients, and higher scores went with poorer function and more health care use (Kroenke et al., 2001).

What are the types of major depressive disorder?

The diagnostic manual classifies the disorder by course, as a single episode or recurrent, by current severity and remission status, and by specifiers describing features of the episode, such as anxious distress, melancholic, atypical or psychotic features, catatonia, peripartum onset and a seasonal pattern (American Psychiatric Association, 2022). Researchers add further constructs, such as treatment-resistant depression and vascular depression, that are not diagnostic categories. Anxious depression went with poorer remission in one large trial, and cognitive deficits are already measurable at a first episode.

Does major depressive disorder affect memory and concentration?

A meta-analysis of studies using the CANTAB battery found moderate deficits in executive function, memory and attention during depression; in remitted patients the executive and attentional deficits remained, while the memory difference was non-significant (Rock et al., 2014). One small study argues that momentary influences during testing, such as well-being, make these deficits look larger than they are. After electroconvulsive therapy, deficits on standardised tests were concentrated in the first few days.

Is major depressive disorder caused by negative thinking?

Not on present evidence. An early review found many aspects of the cognitive theory's description of depressive thinking substantiated, though not the claim that such thinking is especially inaccurate or illogical, and found no strong evidence for its causal claims (Haaga et al., 1991). Later prospective studies show that negative cognitive styles can precede an episode, yet whether they cause it remains an open question.

How well do psychotherapy and antidepressants work?

In a network meta-analysis of 101 studies with 11,910 patients, psychotherapy and medication produced similar response rates, and combining them produced more responders than either alone; the authors judged combined treatment the best choice for moderate depression (Cuijpers et al., 2020). Comparisons with unpublished and regulatory data suggest that published trials overstate both treatments, and whether the drug advantage depends on baseline severity remains disputed. Interpersonal psychotherapy outperformed control conditions in the 16 studies that compared it with a control group, and in young people only fluoxetine beat placebo among the antidepressants compared, while psychotherapy helped modestly.

Can relapse be prevented?

In an individual patient data meta-analysis of 1258 people with recurrent depression, mindfulness-based cognitive therapy lowered the risk of relapse over 60 weeks (hazard ratio 0.69), and the authors judged that it appears efficacious, particularly for those with more pronounced residual symptoms, and made recommendations for addressing remaining uncertainties (Kuyken et al., 2016). Continuing antidepressant medication is the other strategy with strong trial support. In one primary care trial, people who stopped their medication relapsed more often than those who continued it, and in a head-to-head trial mindfulness-based cognitive therapy was not superior to maintenance medication.

What raises the risk of suicide in depression, and where can someone get help?

A systematic review of nineteen studies in depressed samples identified seven factors associated with later suicide, among them a previous suicide attempt, greater severity of depression and hopelessness, and the authors recommended that such factors inform clinical risk assessment (Hawton et al., 2013). Risk factors of this kind mark raised risk but are poor at telling who will act, since across 50 years of studies broad risk factors forecast suicidal thoughts and behaviours only slightly better than chance. Anyone who is thinking about suicide can contact a local crisis line or emergency services for immediate help.

References

Abramson, L. Y., Metalsky, G. I., & Alloy, L. B. (1989). Hopelessness depression: A theory-based subtype of depression. Psychological Review, 96(2), 358-372. https://doi.org/10.1037/0033-295X.96.2.358

Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D. (1978). Learned helplessness in humans: Critique and reformulation. Journal of Abnormal Psychology, 87(1), 49-74. https://doi.org/10.1037/0021-843X.87.1.49

Ahern, E., & Semkovska, M. (2017). Cognitive functioning in the first-episode of major depressive disorder: A systematic review and meta-analysis. Neuropsychology, 31(1), 52-72. https://doi.org/10.1037/neu0000319

Alexopoulos, G. S., Meyers, B. S., Young, R. C., Campbell, S., Silbersweig, D., & Charlson, M. (1997). 'Vascular depression' hypothesis. Archives of General Psychiatry, 54(10), 915-922. https://doi.org/10.1001/archpsyc.1997.01830220033006

Alloy, L. B., & Abramson, L. Y. (1979). Judgment of contingency in depressed and nondepressed students: Sadder but wiser? Journal of Experimental Psychology: General, 108(4), 441-485. https://doi.org/10.1037/0096-3445.108.4.441

Alloy, L. B., Abramson, L. Y., Whitehouse, W. G., Hogan, M. E., Panzarella, C., & Rose, D. T. (2006). Prospective incidence of first onsets and recurrences of depression in individuals at high and low cognitive risk for depression. Journal of Abnormal Psychology, 115(1), 145-156. https://doi.org/10.1037/0021-843X.115.1.145

American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.; DSM-5-TR). American Psychiatric Association Publishing. https://doi.org/10.1176/appi.books.9780890425787

Anand, A., Mathew, S. J., Sanacora, G., Murrough, J. W., Goes, F. S., Altinay, M., Aloysi, A. S., Asghar-Ali, A. A., Barnett, B. S., Chang, L. C., Collins, K. A., Costi, S., Iqbal, S., Jha, M. K., Krishnan, K., Malone, D. A., Nikayin, S., Nissen, S. E., Ostroff, R. B., Reti, I. M., Wilkinson, S. T., Wolski, K., & Hu, B. (2023). Ketamine versus ECT for nonpsychotic treatment-resistant major depression. New England Journal of Medicine, 388(25), 2315-2325. https://doi.org/10.1056/NEJMoa2302399

Andrade, L. H., Alonso, J., Mneimneh, Z., Wells, J. E., Al-Hamzawi, A., Borges, G., Bromet, E., Bruffaerts, R., de Girolamo, G., de Graaf, R., Florescu, S., Gureje, O., Hinkov, H. R., Hu, C., Huang, Y., Hwang, I., Jin, R., Karam, E. G., Kovess-Masfety, V., ... Kessler, R. C. (2014). Barriers to mental health treatment: Results from the WHO World Mental Health surveys. Psychological Medicine, 44(6), 1303-1317. https://doi.org/10.1017/S0033291713001943

Andrews, P. W., & Thomson, J. A., Jr. (2009). The bright side of being blue: Depression as an adaptation for analyzing complex problems. Psychological Review, 116(3), 620-654. https://doi.org/10.1037/a0016242

Baglioni, C., Battagliese, G., Feige, B., Spiegelhalder, K., Nissen, C., Voderholzer, U., Lombardo, C., & Riemann, D. (2011). Insomnia as a predictor of depression: A meta-analytic evaluation of longitudinal epidemiological studies. Journal of Affective Disorders, 135(1-3), 10-19. https://doi.org/10.1016/j.jad.2011.01.011

Baglioni, C., Nanovska, S., Regen, W., Spiegelhalder, K., Feige, B., Nissen, C., Reynolds, C. F., & Riemann, D. (2016). Sleep and mental disorders: A meta-analysis of polysomnographic research. Psychological Bulletin, 142(9), 969-990. https://doi.org/10.1037/bul0000053

Bahji, A., Vazquez, G. H., & Zarate, C. A., Jr. (2021). Comparative efficacy of racemic ketamine and esketamine for depression: A systematic review and meta-analysis. Journal of Affective Disorders, 278, 542-555. https://doi.org/10.1016/j.jad.2020.09.071

Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. Hoeber Medical Division, Harper & Row.

Beck, A. T. (2008). The evolution of the cognitive model of depression and its neurobiological correlates. American Journal of Psychiatry, 165(8), 969-977. https://doi.org/10.1176/appi.ajp.2008.08050721

Beck, A. T., & Bredemeier, K. (2016). A unified model of depression: Integrating clinical, cognitive, biological, and evolutionary perspectives. Clinical Psychological Science, 4(4), 596-619. https://doi.org/10.1177/2167702616628523

Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4(6), 561-571. https://doi.org/10.1001/archpsyc.1961.01710120031004

Border, R., Johnson, E. C., Evans, L. M., et al. (2019). No support for historical candidate gene or candidate gene-by-interaction hypotheses for major depression across multiple large samples. American Journal of Psychiatry, 176(5), 376-387. https://doi.org/10.1176/appi.ajp.2018.18070881

Borsboom, D., & Cramer, A. O. J. (2013). Network analysis: An integrative approach to the structure of psychopathology. Annual Review of Clinical Psychology, 9, 91-121. https://doi.org/10.1146/annurev-clinpsy-050212-185608

Borsboom, D., Fried, E. I., Epskamp, S., Waldorp, L. J., van Borkulo, C. D., van der Maas, H. L. J., & Cramer, A. O. J. (2017). False alarm? A comprehensive reanalysis of "Evidence that psychopathology symptom networks have limited replicability" by Forbes, Wright, Markon, and Krueger (2017). Journal of Abnormal Psychology, 126(7), 989-999. https://doi.org/10.1037/abn0000306

Bromet, E., Andrade, L. H., Hwang, I., Sampson, N. A., Alonso, J., de Girolamo, G., de Graaf, R., Demyttenaere, K., Hu, C., Iwata, N., Karam, A. N., Kaur, J., Kostyuchenko, S., Lépine, J.-P., Levinson, D., Matschinger, H., Medina Mora, M. E., Oakley Browne, M., Posada-Villa, J., Viana, M. C., Williams, D. R., & Kessler, R. C. (2011). Cross-national epidemiology of DSM-IV major depressive episode. BMC Medicine, 9, 90. https://doi.org/10.1186/1741-7015-9-90

Brown, S., Rittenbach, K., Cheung, S., McKean, G., MacMaster, F. P., & Clement, F. (2019). Current and common definitions of treatment-resistant depression: Findings from a systematic review and qualitative interviews. Canadian Journal of Psychiatry, 64(6), 380-387. https://doi.org/10.1177/0706743719828965

Brunoni, A. R., Chaimani, A., Moffa, A. H., Razza, L. B., Gattaz, W. F., Daskalakis, Z. J., & Carvalho, A. F. (2017). Repetitive transcranial magnetic stimulation for the acute treatment of major depressive episodes: A systematic review with network meta-analysis. JAMA Psychiatry, 74(2), 143-152. https://doi.org/10.1001/jamapsychiatry.2016.3644

Buckman, J. E. J., Underwood, A., Clarke, K., Saunders, R., Hollon, S. D., Fearon, P., & Pilling, S. (2018). Risk factors for relapse and recurrence of depression in adults and how they operate: A four-phase systematic review and meta-synthesis. Clinical Psychology Review, 64, 13-38. https://doi.org/10.1016/j.cpr.2018.07.005

Burcusa, S. L., & Iacono, W. G. (2007). Risk for recurrence in depression. Clinical Psychology Review, 27(8), 959-985. https://doi.org/10.1016/j.cpr.2007.02.005

Buus, N., Petersen, A., McPherson, S., Meadows, G., Brand, G., & Ong, B. (2024). The relatives of people with depression: A systematic review and methodological critique of qualitative studies. Family Process, 63(3), 1469-1483. https://doi.org/10.1111/famp.12927

Cambridge, O. R., Knight, M. J., Mills, N., & Baune, B. T. (2018). The clinical relationship between cognitive impairment and psychosocial functioning in major depressive disorder: A systematic review. Psychiatry Research, 269, 157-171. https://doi.org/10.1016/j.psychres.2018.08.033

Caspi, A., Sugden, K., Moffitt, T. E., Taylor, A., Craig, I. W., Harrington, H., McClay, J., Mill, J., Martin, J., Braithwaite, A., & Poulton, R. (2003). Influence of life stress on depression: Moderation by a polymorphism in the 5-HTT gene. Science, 301(5631), 386-389. https://doi.org/10.1126/science.1083968

Chu, C., Buchman-Schmitt, J. M., Stanley, I. H., Hom, M. A., Tucker, R. P., Hagan, C. R., Rogers, M. L., Podlogar, M. C., Chiurliza, B., Ringer, F. B., Michaels, M. S., Patros, C. H. G., & Joiner, T. E. (2017). The interpersonal theory of suicide: A systematic review and meta-analysis of a decade of cross-national research. Psychological Bulletin, 143(12), 1313-1345. https://doi.org/10.1037/bul0000123

Cipriani, A., Furukawa, T. A., Salanti, G., et al. (2018). Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: A systematic review and network meta-analysis. The Lancet, 391(10128), 1357-1366. https://doi.org/10.1016/S0140-6736(17)32802-7

Cipriani, A., Zhou, X., Del Giovane, C., Hetrick, S. E., Qin, B., Whittington, C., Coghill, D., Zhang, Y., Hazell, P., Leucht, S., Cuijpers, P., Pu, J., Cohen, D., Ravindran, A. V., Liu, Y., Michael, K. D., Yang, L., Liu, L., & Xie, P. (2016). Comparative efficacy and tolerability of antidepressants for major depressive disorder in children and adolescents: A network meta-analysis. The Lancet, 388(10047), 881-890. https://doi.org/10.1016/S0140-6736(16)30385-3

Clement, S., Schauman, O., Graham, T., Maggioni, F., Evans-Lacko, S., Bezborodovs, N., Morgan, C., Rüsch, N., Brown, J. S. L., & Thornicroft, G. (2015). What is the impact of mental health-related stigma on help-seeking? A systematic review of quantitative and qualitative studies. Psychological Medicine, 45(1), 11-27. https://doi.org/10.1017/S0033291714000129

Coyne, J. C. (1976). Toward an interactional description of depression. Psychiatry, 39(1), 28-40. https://doi.org/10.1080/00332747.1976.11023874

Cristea, I. A., Kok, R. N., & Cuijpers, P. (2015). Efficacy of cognitive bias modification interventions in anxiety and depression: Meta-analysis. British Journal of Psychiatry, 206(1), 7-16. https://doi.org/10.1192/bjp.bp.114.146761

Crowe, M., Inder, M., & McCall, C. (2023). Experience of antidepressant use and discontinuation: A qualitative synthesis of the evidence. Journal of Psychiatric and Mental Health Nursing, 30(1), 21-34. https://doi.org/10.1111/jpm.12850

Cuijpers, P., Geraedts, A. S., van Oppen, P., Andersson, G., Markowitz, J. C., & van Straten, A. (2011). Interpersonal psychotherapy for depression: A meta-analysis. American Journal of Psychiatry, 168(6), 581-592. https://doi.org/10.1176/appi.ajp.2010.10101411

Cuijpers, P., Karyotaki, E., Reijnders, M., & Ebert, D. D. (2019). Was Eysenck right after all? A reassessment of the effects of psychotherapy for adult depression. Epidemiology and Psychiatric Sciences, 28(1), 21-30. https://doi.org/10.1017/S2045796018000057

Cuijpers, P., Noma, H., Karyotaki, E., Vinkers, C. H., Cipriani, A., & Furukawa, T. A. (2020). A network meta-analysis of the effects of psychotherapies, pharmacotherapies and their combination in the treatment of adult depression. World Psychiatry, 19(1), 92-107. https://doi.org/10.1002/wps.20701

Cuijpers, P., Vogelzangs, N., Twisk, J., Kleiboer, A., Li, J., & Penninx, B. W. (2014). Comprehensive meta-analysis of excess mortality in depression in the general community versus patients with specific illnesses. American Journal of Psychiatry, 171(4), 453-462. https://doi.org/10.1176/appi.ajp.2013.13030325

Culverhouse, R. C., Saccone, N. L., Horton, A. C., Ma, Y., Anstey, K. J., et al. (2018). Collaborative meta-analysis finds no evidence of a strong interaction between stress and 5-HTTLPR genotype contributing to the development of depression. Molecular Psychiatry, 23(1), 133-142. https://doi.org/10.1038/mp.2017.44

Davies, J., & Read, J. (2019). A systematic review into the incidence, severity and duration of antidepressant withdrawal effects: Are guidelines evidence-based? Addictive Behaviors, 97, 111-121. https://doi.org/10.1016/j.addbeh.2018.08.027

Dhingra, K., Boduszek, D., & O'Connor, R. C. (2015). Differentiating suicide attempters from suicide ideators using the Integrated Motivational-Volitional model of suicidal behaviour. Journal of Affective Disorders, 186, 211-218. https://doi.org/10.1016/j.jad.2015.07.007

Dimidjian, S., Hollon, S. D., Dobson, K. S., Schmaling, K. B., Kohlenberg, R. J., Addis, M. E., Gallop, R., McGlinchey, J. B., Markley, D. K., Gollan, J. K., Atkins, D. C., Dunner, D. L., & Jacobson, N. S. (2006). Randomized trial of behavioral activation, cognitive therapy, and antidepressant medication in the acute treatment of adults with major depression. Journal of Consulting and Clinical Psychology, 74(4), 658-670. https://doi.org/10.1037/0022-006X.74.4.658

Dinga, R., Schmaal, L., Penninx, B. W. J. H., van Tol, M. J., Veltman, D. J., van Velzen, L., Mennes, M., van der Wee, N. J. A., & Marquand, A. F. (2019). Evaluating the evidence for biotypes of depression: Methodological replication and extension of Drysdale et al. (2017). NeuroImage: Clinical, 22, 101796. https://doi.org/10.1016/j.nicl.2019.101796

Diniz, B. S., Butters, M. A., Albert, S. M., Dew, M. A., & Reynolds, C. F., III. (2013). Late-life depression and risk of vascular dementia and Alzheimer's disease: Systematic review and meta-analysis of community-based cohort studies. The British Journal of Psychiatry, 202(5), 329-335. https://doi.org/10.1192/bjp.bp.112.118307

Disner, S. G., Beevers, C. G., Haigh, E. A. P., & Beck, A. T. (2011). Neural mechanisms of the cognitive model of depression. Nature Reviews Neuroscience, 12(8), 467-477. https://doi.org/10.1038/nrn3027

Doblyte, S., & Jiménez-Mejías, E. (2017). Understanding help-seeking behavior in depression: A qualitative synthesis of patients' experiences. Qualitative Health Research, 27(1), 100-113. https://doi.org/10.1177/1049732316681282

Dowlati, Y., Herrmann, N., Swardfager, W., Liu, H., Sham, L., Reim, E. K., & Lanctôt, K. L. (2010). A meta-analysis of cytokines in major depression. Biological Psychiatry, 67(5), 446-457. https://doi.org/10.1016/j.biopsych.2009.09.033

Driessen, E., Hollon, S. D., Bockting, C. L. H., Cuijpers, P., & Turner, E. H. (2015). Does publication bias inflate the apparent efficacy of psychological treatment for major depressive disorder? A systematic review and meta-analysis of US National Institutes of Health-funded trials. PLoS ONE, 10(9), e0137864. https://doi.org/10.1371/journal.pone.0137864

Drysdale, A. T., Grosenick, L., Downar, J., Dunlop, K., Mansouri, F., Meng, Y., Fetcho, R. N., Zebley, B., Oathes, D. J., Etkin, A., Schatzberg, A. F., Sudheimer, K., Keller, J., Mayberg, H. S., Gunning, F. M., Alexopoulos, G. S., Fox, M. D., Pascual-Leone, A., Voss, H. U., Casey, B. J., Dubin, M. J., & Liston, C. (2017). Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nature Medicine, 23(1), 28-38. https://doi.org/10.1038/nm.4246

Eckshtain, D., Kuppens, S., Ugueto, A., Ng, M. Y., Vaughn-Coaxum, R., Corteselli, K., & Weisz, J. R. (2020). Meta-analysis: 13-year follow-up of psychotherapy effects on youth depression. Journal of the American Academy of Child & Adolescent Psychiatry, 59(1), 45-63. https://doi.org/10.1016/j.jaac.2019.04.002

Evans, V. C., Iverson, G. L., Yatham, L. N., & Lam, R. W. (2014). The relationship between neurocognitive and psychosocial functioning in major depressive disorder: A systematic review. The Journal of Clinical Psychiatry, 75(12), 1359-1370. https://doi.org/10.4088/JCP.13r08939

Everaert, J., Podina, I. R., & Koster, E. H. W. (2017). A comprehensive meta-analysis of interpretation biases in depression. Clinical Psychology Review, 58, 33-48. https://doi.org/10.1016/j.cpr.2017.09.005

Fava, M., Rush, A. J., Alpert, J. E., Balasubramani, G. K., Wisniewski, S. R., Carmin, C. N., Biggs, M. M., Zisook, S., Leuchter, A., Howland, R., Warden, D., & Trivedi, M. H. (2008). Difference in treatment outcome in outpatients with anxious versus nonanxious depression: A STAR*D report. American Journal of Psychiatry, 165(3), 342-351. https://doi.org/10.1176/appi.ajp.2007.06111868

Feighner, J. P., Robins, E., Guze, S. B., Woodruff, R. A., Jr., Winokur, G., & Munoz, R. (1972). Diagnostic criteria for use in psychiatric research. Archives of General Psychiatry, 26(1), 57-63. https://doi.org/10.1001/archpsyc.1972.01750190059011

Fodor, L. A., Georgescu, R., Cuijpers, P., Szamoskozi, Ş., David, D., Furukawa, T. A., & Cristea, I. A. (2020). Efficacy of cognitive bias modification interventions in anxiety and depressive disorders: A systematic review and network meta-analysis. The Lancet Psychiatry, 7(6), 506-514. https://doi.org/10.1016/S2215-0366(20)30130-9

Forbes, M. K., Wright, A. G. C., Markon, K. E., & Krueger, R. F. (2017). Evidence that psychopathology symptom networks have limited replicability. Journal of Abnormal Psychology, 126(7), 969-988. https://doi.org/10.1037/abn0000276

Fournier, J. C., DeRubeis, R. J., Hollon, S. D., Dimidjian, S., Amsterdam, J. D., Shelton, R. C., & Fawcett, J. (2010). Antidepressant drug effects and depression severity: A patient-level meta-analysis. JAMA, 303(1), 47-53. https://doi.org/10.1001/jama.2009.1943

Franklin, J. C., Ribeiro, J. D., Fox, K. R., Bentley, K. H., Kleiman, E. M., Huang, X., Musacchio, K. M., Jaroszewski, A. C., Chang, B. P., & Nock, M. K. (2017). Risk factors for suicidal thoughts and behaviors: A meta-analysis of 50 years of research. Psychological Bulletin, 143(2), 187-232. https://doi.org/10.1037/bul0000084

Fried, E. I., & Nesse, R. M. (2015). Depression sum-scores don't add up: Why analyzing specific depression symptoms is essential. BMC Medicine, 13, 72. https://doi.org/10.1186/s12916-015-0325-4

Furukawa, T. A., Maruo, K., Noma, H., Tanaka, S., Imai, H., Shinohara, K., Ikeda, K., Yamawaki, S., Levine, S. Z., Goldberg, Y., Leucht, S., & Cipriani, A. (2018). Initial severity of major depression and efficacy of new generation antidepressants: Individual participant data meta-analysis. Acta Psychiatrica Scandinavica, 137(6), 450-458. https://doi.org/10.1111/acps.12886

GBD 2019 Mental Disorders Collaborators. (2022). Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990-2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet Psychiatry, 9(2), 137-150. https://doi.org/10.1016/S2215-0366(21)00395-3

Goodman, S. H., Rouse, M. H., Connell, A. M., Broth, M. R., Hall, C. M., & Heyward, D. (2011). Maternal depression and child psychopathology: A meta-analytic review. Clinical Child and Family Psychology Review, 14(1), 1-27. https://doi.org/10.1007/s10567-010-0080-1

Gotlib, I. H., & Joormann, J. (2010). Cognition and depression: Current status and future directions. Annual Review of Clinical Psychology, 6, 285-312. https://doi.org/10.1146/annurev.clinpsy.121208.131305

Haaga, D. A. F., Dyck, M. J., & Ernst, D. (1991). Empirical status of cognitive theory of depression. Psychological Bulletin, 110(2), 215-236. https://doi.org/10.1037/0033-2909.110.2.215

Hackett, M. L., & Pickles, K. (2014). Part I: Frequency of depression after stroke: An updated systematic review and meta-analysis of observational studies. International Journal of Stroke, 9(8), 1017-1025. https://doi.org/10.1111/ijs.12357

Halahakoon, D. C., Kieslich, K., O'Driscoll, C., Nair, A., Lewis, G., & Roiser, J. P. (2020). Reward-processing behavior in depressed participants relative to healthy volunteers: A systematic review and meta-analysis. JAMA Psychiatry, 77(12), 1286-1295. https://doi.org/10.1001/jamapsychiatry.2020.2139

Hamilton, J. P., Farmer, M., Fogelman, P., & Gotlib, I. H. (2015). Depressive rumination, the default-mode network, and the dark matter of clinical neuroscience. Biological Psychiatry, 78(4), 224-230. https://doi.org/10.1016/j.biopsych.2015.02.020

Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery, and Psychiatry, 23(1), 56-62. https://doi.org/10.1136/jnnp.23.1.56

Hammen, C. (1991). Generation of stress in the course of unipolar depression. Journal of Abnormal Psychology, 100(4), 555-561. https://doi.org/10.1037/0021-843X.100.4.555

Hardeveld, F., Spijker, J., De Graaf, R., Nolen, W. A., & Beekman, A. T. F. (2010). Prevalence and predictors of recurrence of major depressive disorder in the adult population. Acta Psychiatrica Scandinavica, 122(3), 184-191. https://doi.org/10.1111/j.1600-0447.2009.01519.x

Harmer, C. J., Goodwin, G. M., & Cowen, P. J. (2009). Why do antidepressants take so long to work? A cognitive neuropsychological model of antidepressant drug action. British Journal of Psychiatry, 195(2), 102-108. https://doi.org/10.1192/bjp.bp.108.051193

Hasin, D. S., Sarvet, A. L., Meyers, J. L., Saha, T. D., Ruan, W. J., Stohl, M., & Grant, B. F. (2018). Epidemiology of adult DSM-5 major depressive disorder and its specifiers in the United States. JAMA Psychiatry, 75(4), 336-346. https://doi.org/10.1001/jamapsychiatry.2017.4602

Haslam, N., Holland, E., & Kuppens, P. (2012). Categories versus dimensions in personality and psychopathology: A quantitative review of taxometric research. Psychological Medicine, 42(5), 903-920. https://doi.org/10.1017/S0033291711001966

Hawton, K., Casañas i Comabella, C., Haw, C., & Saunders, K. (2013). Risk factors for suicide in individuals with depression: A systematic review. Journal of Affective Disorders, 147(1-3), 17-28. https://doi.org/10.1016/j.jad.2013.01.004

Henssler, J., Schmidt, Y., Schmidt, U., Schwarzer, G., Bschor, T., & Baethge, C. (2024). Incidence of antidepressant discontinuation symptoms: A systematic review and meta-analysis. The Lancet Psychiatry, 11(7), 526-535. https://doi.org/10.1016/S2215-0366(24)00133-0

Hertenstein, E., Trinca, E., Wunderlin, M., Schneider, C. L., Züst, M. A., Fehér, K. D., Su, T., van Straten, A., Berger, T., Baglioni, C., Johann, A., Spiegelhalder, K., Riemann, D., Feige, B., & Nissen, C. (2022). Cognitive behavioral therapy for insomnia in patients with mental disorders and comorbid insomnia: A systematic review and meta-analysis. Sleep Medicine Reviews, 62, 101597. https://doi.org/10.1016/j.smrv.2022.101597

Hirshbein, L. D. (2009). Gender, age, and diagnosis: The rise and fall of involutional melancholia in American psychiatry, 1900-1980. Bulletin of the History of Medicine, 83(4), 710-745. https://doi.org/10.1353/bhm.0.0290

Howard, D. M., Adams, M. J., Clarke, T.-K., et al. (2019). Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions. Nature Neuroscience, 22(3), 343-352. https://doi.org/10.1038/s41593-018-0326-7

Janssen, N. P., Hendriks, G.-J., Baranelli, C. T., Lucassen, P., Oude Voshaar, R., Spijker, J., & Huibers, M. J. H. (2021). How does behavioural activation work? A systematic review of the evidence on potential mediators. Psychotherapy and Psychosomatics, 90(2), 85-93. https://doi.org/10.1159/000509820

Jauhar, S., Arnone, D., Baldwin, D. S., Bloomfield, M., Browning, M., Cleare, A. J., ... Cowen, P. J. (2023). A leaky umbrella has little value: Evidence clearly indicates the serotonin system is implicated in depression. Molecular Psychiatry, 28(8), 3149-3152. https://doi.org/10.1038/s41380-023-02095-y

Joormann, J., & Gotlib, I. H. (2010). Emotion regulation in depression: Relation to cognitive inhibition. Cognition and Emotion, 24(2), 281-298. https://doi.org/10.1080/02699930903407948

Kaiser, B. N., Haroz, E. E., Kohrt, B. A., Bolton, P. A., Bass, J. K., & Hinton, D. E. (2015). "Thinking too much": A systematic review of a common idiom of distress. Social Science & Medicine, 147, 170-183. https://doi.org/10.1016/j.socscimed.2015.10.044

Kaiser, R. H., Andrews-Hanna, J. R., Wager, T. D., & Pizzagalli, D. A. (2015). Large-scale network dysfunction in major depressive disorder: A meta-analysis of resting-state functional connectivity. JAMA Psychiatry, 72(6), 603-611. https://doi.org/10.1001/jamapsychiatry.2015.0071

Kendler, K. S. (2016). The phenomenology of major depression and the representativeness and nature of DSM criteria. American Journal of Psychiatry, 173(8), 771-780. https://doi.org/10.1176/appi.ajp.2016.15121509

Kendler, K. S., Gardner, C. O., Gatz, M., & Pedersen, N. L. (2007). The sources of co-morbidity between major depression and generalized anxiety disorder in a Swedish national twin sample. Psychological Medicine, 37(3), 453-462. https://doi.org/10.1017/S0033291706009135

Kendler, K. S., Karkowski, L. M., & Prescott, C. A. (1999). Causal relationship between stressful life events and the onset of major depression. American Journal of Psychiatry, 156(6), 837-841. https://doi.org/10.1176/ajp.156.6.837

Keren, H., O'Callaghan, G., Vidal-Ribas, P., Buzzell, G. A., Brotman, M. A., Leibenluft, E., Pan, P. M., Meffert, L., Kaiser, A., Wolke, S., Pine, D. S., & Stringaris, A. (2018). Reward processing in depression: A conceptual and meta-analytic review across fMRI and EEG studies. American Journal of Psychiatry, 175(11), 1111-1120. https://doi.org/10.1176/appi.ajp.2018.17101124

Kessler, R. C., Berglund, P., Demler, O., Jin, R., Koretz, D., Merikangas, K. R., Rush, A. J., Walters, E. E., & Wang, P. S. (2003). The epidemiology of major depressive disorder: Results from the National Comorbidity Survey Replication (NCS-R). JAMA, 289(23), 3095-3105. https://doi.org/10.1001/jama.289.23.3095

Kirsch, I., Deacon, B. J., Huedo-Medina, T. B., Scoboria, A., Moore, T. J., & Johnson, B. T. (2008). Initial severity and antidepressant benefits: A meta-analysis of data submitted to the Food and Drug Administration. PLoS Medicine, 5(2), e45. https://doi.org/10.1371/journal.pmed.0050045

Kleinman, A. (1982). Neurasthenia and depression: A study of somatization and culture in China. Culture, Medicine and Psychiatry, 6(2), 117-190. https://doi.org/10.1007/BF00051427

Koukopoulos, A., & Sani, G. (2014). DSM-5 criteria for depression with mixed features: A farewell to mixed depression. Acta Psychiatrica Scandinavica, 129(1), 4-16. https://doi.org/10.1111/acps.12140

Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9: Validity of a brief depression severity measure. Journal of General Internal Medicine, 16(9), 606-613. https://doi.org/10.1046/j.1525-1497.2001.016009606.x

Kuehner, C. (2017). Why is depression more common among women than among men? The Lancet Psychiatry, 4(2), 146-158. https://doi.org/10.1016/S2215-0366(16)30263-2

Kuyken, W., Hayes, R., Barrett, B., Byng, R., Dalgleish, T., Kessler, D., Lewis, G., Watkins, E., et al. (2015). Effectiveness and cost-effectiveness of mindfulness-based cognitive therapy compared with maintenance antidepressant treatment in the prevention of depressive relapse or recurrence (PREVENT): A randomised controlled trial. The Lancet, 386(9988), 63-73. https://doi.org/10.1016/S0140-6736(14)62222-4

Kuyken, W., Warren, F. C., Taylor, R. S., Whalley, B., Crane, C., Bondolfi, G., Hayes, R., Huijbers, M., Ma, H., Schweizer, S., Segal, Z., Speckens, A., Teasdale, J. D., Van Heeringen, K., Williams, M., Byford, S., Byng, R., & Dalgleish, T. (2016). Efficacy of mindfulness-based cognitive therapy in prevention of depressive relapse: An individual patient data meta-analysis from randomized trials. JAMA Psychiatry, 73(6), 565-574. https://doi.org/10.1001/jamapsychiatry.2016.0076

Lee, R. S. C., Hermens, D. F., Porter, M. A., & Redoblado-Hodge, M. A. (2012). A meta-analysis of cognitive deficits in first-episode Major Depressive Disorder. Journal of Affective Disorders, 140(2), 113-124. https://doi.org/10.1016/j.jad.2011.10.023

Lee, Y. Y., Stockings, E. A., Harris, M. G., Doi, S. A. R., Page, I. S., Davidson, S. K., & Barendregt, J. J. (2019). The risk of developing major depression among individuals with subthreshold depression: A systematic review and meta-analysis of longitudinal cohort studies. Psychological Medicine, 49(1), 92-102. https://doi.org/10.1017/S0033291718000557

Lemmens, L. H. J. M., Müller, V. N. L. S., Arntz, A., & Huibers, M. J. H. (2016). Mechanisms of change in psychotherapy for depression: An empirical update and evaluation of research aimed at identifying psychological mediators. Clinical Psychology Review, 50, 95-107. https://doi.org/10.1016/j.cpr.2016.09.004

LeMoult, J., & Gotlib, I. H. (2019). Depression: A cognitive perspective. Clinical Psychology Review, 69, 51-66. https://doi.org/10.1016/j.cpr.2018.06.008

Lewis, G., Marston, L., Duffy, L., Freemantle, N., Gilbody, S., Hunter, R., Kendrick, T., Kessler, D., Mangin, D., King, M., Lanham, P., Moore, M., Nazareth, I., Wiles, N., Bacon, F., Bird, M., Brabyn, S., Burns, A., Clarke, C. S., Hunt, A., Pervin, J., & Lewis, G. (2021). Maintenance or discontinuation of antidepressants in primary care. New England Journal of Medicine, 385(14), 1257-1267. https://doi.org/10.1056/NEJMoa2106356

Liu, R. T., & Alloy, L. B. (2010). Stress generation in depression: A systematic review of the empirical literature and recommendations for future study. Clinical Psychology Review, 30(5), 582-593. https://doi.org/10.1016/j.cpr.2010.04.010

Liu, R. T., Kleiman, E. M., Nestor, B. A., & Cheek, S. M. (2015). The hopelessness theory of depression: A quarter century in review. Clinical Psychology: Science and Practice, 22(4), 345-365. https://doi.org/10.1111/cpsp.12125

Longmore, R. J., & Worrell, M. (2007). Do we need to challenge thoughts in cognitive behavior therapy? Clinical Psychology Review, 27(2), 173-187. https://doi.org/10.1016/j.cpr.2006.08.001

Lorant, V., Deliège, D., Eaton, W., Robert, A., Philippot, P., & Ansseau, M. (2003). Socioeconomic inequalities in depression: A meta-analysis. American Journal of Epidemiology, 157(2), 98-112. https://doi.org/10.1093/aje/kwf182

Lorenzo-Luaces, L., German, R. E., & DeRubeis, R. J. (2015). It's complicated: The relation between cognitive change procedures, cognitive change, and symptom change in cognitive therapy for depression. Clinical Psychology Review, 41, 3-15. https://doi.org/10.1016/j.cpr.2014.12.003

Lund, C., Brooke-Sumner, C., Baingana, F., Baron, E. C., Breuer, E., Chandra, P., Haushofer, J., Herrman, H., Jordans, M., Kieling, C., Medina-Mora, M. E., Morgan, E., Omigbodun, O., Tol, W., Patel, V., & Saxena, S. (2018). Social determinants of mental disorders and the Sustainable Development Goals: A systematic review of reviews. The Lancet Psychiatry, 5(4), 357-369. https://doi.org/10.1016/S2215-0366(18)30060-9

Maier, S. F., & Seligman, M. E. P. (2016). Learned helplessness at fifty: Insights from neuroscience. Psychological Review, 123(4), 349-367. https://doi.org/10.1037/rev0000033

Malhi, G. S., & Mann, J. J. (2018). Depression. The Lancet, 392(10161), 2299-2312. https://doi.org/10.1016/S0140-6736(18)31948-2

Marwaha, S., Palmer, E., Suppes, T., Cons, E., Young, A. H., & Upthegrove, R. (2023). Novel and emerging treatments for major depression. The Lancet, 401(10371), 141-153. https://doi.org/10.1016/S0140-6736(22)02080-3

Matt, G. E., Vázquez, C., & Campbell, W. K. (1992). Mood-congruent recall of affectively toned stimuli: A meta-analytic review. Clinical Psychology Review, 12(2), 227-255. https://doi.org/10.1016/0272-7358(92)90116-P

McDermott, L. M., & Ebmeier, K. P. (2009). A meta-analysis of depression severity and cognitive function. Journal of Affective Disorders, 119(1-3), 1-8. https://doi.org/10.1016/j.jad.2009.04.022

McIntyre, R. S., Alsuwaidan, M., Baune, B. T., Berk, M., Demyttenaere, K., Goldberg, J. F., Gorwood, P., Ho, R., Kasper, S., Kennedy, S. H., Ly-Uson, J., Mansur, R. B., McAllister-Williams, R. H., Murrough, J. W., Nemeroff, C. B., Nierenberg, A. A., Rosenblat, J. D., Sanacora, G., Schatzberg, A. F., Shelton, R., Stahl, S. M., Trivedi, M. H., Vieta, E., Vinberg, M., Williams, N., Young, A. H., & Maj, M. (2023). Treatment-resistant depression: Definition, prevalence, detection, management, and investigational interventions. World Psychiatry, 22(3), 394-412. https://doi.org/10.1002/wps.21120

Miller, A. H., Maletic, V., & Raison, C. L. (2009). Inflammation and its discontents: The role of cytokines in the pathophysiology of major depression. Biological Psychiatry, 65(9), 732-741. https://doi.org/10.1016/j.biopsych.2008.11.029

Mitter, P., De Crescenzo, F., Loo Yong Kee, K., Xia, J., Roberts, S., Chi, W., Kurtulmus, A., Kyle, S. D., Geddes, J. R., & Cipriani, A. (2022). Sleep deprivation as a treatment for major depressive episodes: A systematic review and meta-analysis. Sleep Medicine Reviews, 64, 101647. https://doi.org/10.1016/j.smrv.2022.101647

Molendijk, M. L., Spinhoven, P., Polak, M., Bus, B. A. A., Penninx, B. W. J. H., & Elzinga, B. M. (2014). Serum BDNF concentrations as peripheral manifestations of depression: Evidence from a systematic review and meta-analyses on 179 associations (N=9484). Molecular Psychiatry, 19(7), 791-800. https://doi.org/10.1038/mp.2013.105

Moncrieff, J., Cooper, R. E., Stockmann, T., Amendola, S., Hengartner, M. P., & Horowitz, M. A. (2022). The serotonin theory of depression: A systematic umbrella review of the evidence. Molecular Psychiatry, 28(8), 3243-3256. https://doi.org/10.1038/s41380-022-01661-0

Monroe, S. M., & Harkness, K. L. (2005). Life stress, the kindling hypothesis, and the recurrence of depression: Considerations from a life stress perspective. Psychological Review, 112(2), 417-445. https://doi.org/10.1037/0033-295X.112.2.417

Moore, M. T., & Fresco, D. M. (2012). Depressive realism: A meta-analytic review. Clinical Psychology Review, 32(6), 496-509. https://doi.org/10.1016/j.cpr.2012.05.004

Moritz, S., Xie, J., Penney, D., Bihl, L., Hlubek, N., Elmers, J., Beblo, T., & Hottenrott, B. (2023). The magnitude of neurocognitive impairment is overestimated in depression: The role of motivation, debilitating momentary influences, and the overreliance on mean differences. Psychological Medicine, 53(7), 2820-2830. https://doi.org/10.1017/S0033291721004785

Moussavi, S., Chatterji, S., Verdes, E., Tandon, A., Patel, V., & Ustun, B. (2007). Depression, chronic diseases, and decrements in health: Results from the World Health Surveys. The Lancet, 370(9590), 851-858. https://doi.org/10.1016/S0140-6736(07)61415-9

Munkholm, K., Paludan-Müller, A. S., & Boesen, K. (2019). Considering the methodological limitations in the evidence base of antidepressants for depression: A reanalysis of a network meta-analysis. BMJ Open, 9(6), e024886. https://doi.org/10.1136/bmjopen-2018-024886

Na, K.-S., Kang, J. M., & Cho, S.-E. (2021). Prevalence of DSM-5 mixed features: A meta-analysis and systematic review. Journal of Affective Disorders, 282, 203-210. https://doi.org/10.1016/j.jad.2020.12.149

Nanni, V., Uher, R., & Danese, A. (2012). Childhood maltreatment predicts unfavorable course of illness and treatment outcome in depression: A meta-analysis. American Journal of Psychiatry, 169(2), 141-151. https://doi.org/10.1176/appi.ajp.2011.11020335

Nesse, R. M. (2000). Is depression an adaptation? Archives of General Psychiatry, 57(1), 14-20. https://doi.org/10.1001/archpsyc.57.1.14

Ng, T. H., Alloy, L. B., & Smith, D. V. (2019). Meta-analysis of reward processing in major depressive disorder reveals distinct abnormalities within the reward circuit. Translational Psychiatry, 9(1), 293. https://doi.org/10.1038/s41398-019-0644-x

Nolen-Hoeksema, S. (1991). Responses to depression and their effects on the duration of depressive episodes. Journal of Abnormal Psychology, 100(4), 569-582. https://doi.org/10.1037/0021-843X.100.4.569

Nuij, C., van Ballegooijen, W., de Beurs, D., Juniar, D., Erlangsen, A., Portzky, G., O'Connor, R. C., Smit, J. H., Kerkhof, A., & Riper, H. (2021). Safety planning-type interventions for suicide prevention: Meta-analysis. The British Journal of Psychiatry, 219(2), 419-426. https://doi.org/10.1192/bjp.2021.50

O'Connor, R. C., & Kirtley, O. J. (2018). The integrated motivational-volitional model of suicidal behaviour. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1754), 20170268. https://doi.org/10.1098/rstb.2017.0268

Ormel, J., Oldehinkel, A. J., & Vollebergh, W. (2004). Vulnerability before, during, and after a major depressive episode: A 3-wave population-based study. Archives of General Psychiatry, 61(10), 990-996. https://doi.org/10.1001/archpsyc.61.10.990

Osimo, E. F., Baxter, L. J., Lewis, G., Jones, P. B., & Khandaker, G. M. (2019). Prevalence of low-grade inflammation in depression: A systematic review and meta-analysis of CRP levels. Psychological Medicine, 49(12), 1958-1970. https://doi.org/10.1017/S0033291719001454

Otte, C., Gold, S. M., Penninx, B. W., Pariante, C. M., Etkin, A., Fava, M., Mohr, D. C., & Schatzberg, A. F. (2016). Major depressive disorder. Nature Reviews Disease Primers, 2, 16065. https://doi.org/10.1038/nrdp.2016.65

Pariante, C. M., & Lightman, S. L. (2008). The HPA axis in major depression: Classical theories and new developments. Trends in Neurosciences, 31(9), 464-468. https://doi.org/10.1016/j.tins.2008.06.006

Parker, G., Fink, M., Shorter, E., Taylor, M. A., Akiskal, H., Berrios, G., Bolwig, T., Brown, W. A., Carroll, B., Healy, D., Klein, D. F., Koukopoulos, A., Michels, R., Paris, J., Rubin, R. T., Spitzer, R., & Swartz, C. (2010). Issues for DSM-5: Whither melancholia? The case for its classification as a distinct mood disorder. American Journal of Psychiatry, 167(7), 745-747. https://doi.org/10.1176/appi.ajp.2010.09101525

Peckham, A. D., McHugh, R. K., & Otto, M. W. (2010). A meta-analysis of the magnitude of biased attention in depression. Depression and Anxiety, 27(12), 1135-1142. https://doi.org/10.1002/da.20755

Pigott, H. E., Kim, T., Xu, C., Kirsch, I., & Amsterdam, J. (2023). What are the treatment remission, response and extent of improvement rates after up to four trials of antidepressant therapies in real-world depressed patients? A reanalysis of the STAR*D study's patient-level data with fidelity to the original research protocol. BMJ Open, 13(7), e063095. https://doi.org/10.1136/bmjopen-2022-063095

Pizzagalli, D. A., Holmes, A. J., Dillon, D. G., Goetz, E. L., Birk, J. L., Bogdan, R., Dougherty, D. D., Iosifescu, D. V., Rauch, S. L., & Fava, M. (2009). Reduced caudate and nucleus accumbens response to rewards in unmedicated individuals with major depressive disorder. American Journal of Psychiatry, 166(6), 702-710. https://doi.org/10.1176/appi.ajp.2008.08081201

Pjrek, E., Friedrich, M.-E., Cambioli, L., Dold, M., Jäger, F., Komorowski, A., Lanzenberger, R., Kasper, S., & Winkler, D. (2020). The efficacy of light therapy in the treatment of seasonal affective disorder: A meta-analysis of randomized controlled trials. Psychotherapy and Psychosomatics, 89(1), 17-24. https://doi.org/10.1159/000502891

Popova, V., Daly, E. J., Trivedi, M., et al. (2019). Efficacy and safety of flexibly dosed esketamine nasal spray combined with a newly initiated oral antidepressant in treatment-resistant depression: A randomized double-blind active-controlled study. American Journal of Psychiatry, 176(6), 428-438. https://doi.org/10.1176/appi.ajp.2019.19020172

Post, R. M. (1992). Transduction of psychosocial stress into the neurobiology of recurrent affective disorder. American Journal of Psychiatry, 149(8), 999-1010. https://doi.org/10.1176/ajp.149.8.999

Read, J., Cartwright, C., & Gibson, K. (2014). Adverse emotional and interpersonal effects reported by 1829 New Zealanders while taking antidepressants. Psychiatry Research, 216(1), 67-73. https://doi.org/10.1016/j.psychres.2014.01.042

Reilly, T. J., Patel, S., Unachukwu, I. C., Knox, C.-L., Wilson, C. A., Craig, M. C., Schmalenberger, K. M., Eisenlohr-Moul, T. A., & Cullen, A. E. (2024). The prevalence of premenstrual dysphoric disorder: Systematic review and meta-analysis. Journal of Affective Disorders, 349, 534-540. https://doi.org/10.1016/j.jad.2024.01.066

Ribeiro, J. D., Huang, X., Fox, K. R., & Franklin, J. C. (2018). Depression and hopelessness as risk factors for suicide ideation, attempts and death: Meta-analysis of longitudinal studies. British Journal of Psychiatry, 212(5), 279-286. https://doi.org/10.1192/bjp.2018.27

Riemann, D., Krone, L. B., Wulff, K., & Nissen, C. (2020). Sleep, insomnia, and depression. Neuropsychopharmacology, 45(1), 74-89. https://doi.org/10.1038/s41386-019-0411-y

Rock, P. L., Roiser, J. P., Riedel, W. J., & Blackwell, A. D. (2014). Cognitive impairment in depression: A systematic review and meta-analysis. Psychological Medicine, 44(10), 2029-2040. https://doi.org/10.1017/S0033291713002535

Rodebaugh, T. L., Scullin, R. B., Langer, J. K., Dixon, D. J., Huppert, J. D., Bernstein, A., Zvielli, A., & Lenze, E. J. (2016). Unreliability as a threat to understanding psychopathology: The cautionary tale of attentional bias. Journal of Abnormal Psychology, 125(6), 840-851. https://doi.org/10.1037/abn0000184

Rosenthal, N. E., Sack, D. A., Gillin, J. C., Lewy, A. J., Goodwin, F. K., Davenport, Y., Mueller, P. S., Newsome, D. A., & Wehr, T. A. (1984). Seasonal affective disorder: A description of the syndrome and preliminary findings with light therapy. Archives of General Psychiatry, 41(1), 72-80. https://doi.org/10.1001/archpsyc.1984.01790120076010

Rush, A. J., Trivedi, M. H., Wisniewski, S. R., Nierenberg, A. A., Stewart, J. W., Warden, D., Niederehe, G., Thase, M. E., Lavori, P. W., Lebowitz, B. D., McGrath, P. J., Rosenbaum, J. F., Sackeim, H. A., Kupfer, D. J., Luther, J., & Fava, M. (2006). Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: A STAR*D report. American Journal of Psychiatry, 163(11), 1905-1917. https://doi.org/10.1176/ajp.2006.163.11.1905

Ryder, A. G., Yang, J., Zhu, X., Yao, S., Yi, J., Heine, S. J., & Bagby, R. M. (2008). The cultural shaping of depression: Somatic symptoms in China, psychological symptoms in North America? Journal of Abnormal Psychology, 117(2), 300-313. https://doi.org/10.1037/0021-843X.117.2.300

Salk, R. H., Hyde, J. S., & Abramson, L. Y. (2017). Gender differences in depression in representative national samples: Meta-analyses of diagnoses and symptoms. Psychological Bulletin, 143(8), 783-822. https://doi.org/10.1037/bul0000102

Samamé, C., Szmulewicz, A. G., Valerio, M. P., Martino, D. J., & Strejilevich, S. A. (2017). Are major depression and bipolar disorder neuropsychologically distinct? A meta-analysis of comparative studies. European Psychiatry, 39, 17-26. https://doi.org/10.1016/j.eurpsy.2016.06.002

Scher, C. D., Ingram, R. E., & Segal, Z. V. (2005). Cognitive reactivity and vulnerability: Empirical evaluation of construct activation and cognitive diatheses in unipolar depression. Clinical Psychology Review, 25(4), 487-510. https://doi.org/10.1016/j.cpr.2005.01.005

Schramm, E., Klein, D. N., Elsaesser, M., Furukawa, T. A., & Domschke, K. (2020). Review of dysthymia and persistent depressive disorder: History, correlates, and clinical implications. The Lancet Psychiatry, 7(9), 801-812. https://doi.org/10.1016/S2215-0366(20)30099-7

Segal, Z. V., Kennedy, S., Gemar, M., Hood, K., Pedersen, R., & Buis, T. (2006). Cognitive reactivity to sad mood provocation and the prediction of depressive relapse. Archives of General Psychiatry, 63(7), 749-755. https://doi.org/10.1001/archpsyc.63.7.749

Semkovska, M., & McLoughlin, D. M. (2010). Objective cognitive performance associated with electroconvulsive therapy for depression: A systematic review and meta-analysis. Biological Psychiatry, 68(6), 568-577. https://doi.org/10.1016/j.biopsych.2010.06.009

Semkovska, M., Quinlivan, L., O'Grady, T., Johnson, R., Collins, A., O'Connor, J., Knittle, H., Ahern, E., & Gload, T. (2019). Cognitive function following a major depressive episode: A systematic review and meta-analysis. The Lancet Psychiatry, 6(10), 851-861. https://doi.org/10.1016/S2215-0366(19)30291-3

Singh-Manoux, A., Dugravot, A., Fournier, A., Abell, J., Ebmeier, K., Kivimäki, M., & Sabia, S. (2017). Trajectories of depressive symptoms before diagnosis of dementia: A 28-year follow-up study. JAMA Psychiatry, 74(7), 712-718. https://doi.org/10.1001/jamapsychiatry.2017.0660

Slavich, G. M., & Irwin, M. R. (2014). From stress to inflammation and major depressive disorder: A social signal transduction theory of depression. Psychological Bulletin, 140(3), 774-815. https://doi.org/10.1037/a0035302

Smith, J. A., & Rhodes, J. E. (2015). Being depleted and being shaken: An interpretative phenomenological analysis of the experiential features of a first episode of depression. Psychology and Psychotherapy: Theory, Research and Practice, 88(2), 197-209. https://doi.org/10.1111/papt.12034

Snyder, H. R. (2013). Major depressive disorder is associated with broad impairments on neuropsychological measures of executive function: A meta-analysis and review. Psychological Bulletin, 139(1), 81-132. https://doi.org/10.1037/a0028727

Solmi, M., Pigato, G. G., Roiter, B., Guaglianone, A., Martini, L., Fornaro, M., Monaco, F., Carvalho, A. F., Stubbs, B., Veronese, N., & Correll, C. U. (2018). Prevalence of catatonia and its moderators in clinical samples: Results from a meta-analysis and meta-regression analysis. Schizophrenia Bulletin, 44(5), 1133-1150. https://doi.org/10.1093/schbul/sbx157

Spielmans, G. I., Berman, M. I., Linardatos, E., Rosenlicht, N. Z., Perry, A., & Tsai, A. C. (2013). Adjunctive atypical antipsychotic treatment for major depressive disorder: A meta-analysis of depression, quality of life, and safety outcomes. PLoS Medicine, 10(3), e1001403. https://doi.org/10.1371/journal.pmed.1001403

Spitzer, R. L., Endicott, J., & Robins, E. (1978). Research diagnostic criteria: Rationale and reliability. Archives of General Psychiatry, 35(6), 773-782. https://doi.org/10.1001/archpsyc.1978.01770300115013

Stanley, B., Brown, G. K., Brenner, L. A., Galfalvy, H. C., Currier, G. W., Knox, K. L., Chaudhury, S. R., Bush, A. L., & Green, K. L. (2018). Comparison of the safety planning intervention with follow-up vs usual care of suicidal patients treated in the emergency department. JAMA Psychiatry, 75(9), 894-900. https://doi.org/10.1001/jamapsychiatry.2018.1776

Starr, L. R., & Davila, J. (2008). Excessive reassurance seeking, depression, and interpersonal rejection: A meta-analytic review. Journal of Abnormal Psychology, 117(4), 762-775. https://doi.org/10.1037/a0013866

Stetler, C., & Miller, G. E. (2011). Depression and hypothalamic-pituitary-adrenal activation: A quantitative summary of four decades of research. Psychosomatic Medicine, 73(2), 114-126. https://doi.org/10.1097/PSY.0b013e31820ad12b

Stone, M. B., Yaseen, Z. S., Miller, B. J., Richardville, K., Kalaria, S. N., & Kirsch, I. (2022). Response to acute monotherapy for major depressive disorder in randomized, placebo controlled trials submitted to the US Food and Drug Administration: Individual participant data analysis. BMJ, 378, e067606. https://doi.org/10.1136/bmj-2021-067606

Sullivan, P. F., Neale, M. C., & Kendler, K. S. (2000). Genetic epidemiology of major depression: Review and meta-analysis. American Journal of Psychiatry, 157(10), 1552-1562. https://doi.org/10.1176/appi.ajp.157.10.1552

Taylor, W. D., Aizenstein, H. J., & Alexopoulos, G. S. (2013). The vascular depression hypothesis: Mechanisms linking vascular disease with depression. Molecular Psychiatry, 18(9), 963-974. https://doi.org/10.1038/mp.2013.20

Telford, C., McCarthy-Jones, S., Corcoran, R., & Rowse, G. (2012). Experience sampling methodology studies of depression: The state of the art. Psychological Medicine, 42(6), 1119-1129. https://doi.org/10.1017/S0033291711002200

Thapar, A., Eyre, O., Patel, V., & Brent, D. (2022). Depression in young people. The Lancet, 400(10352), 617-631. https://doi.org/10.1016/S0140-6736(22)01012-1

Traffanstedt, M. K., Mehta, S., & LoBello, S. G. (2016). Major depression with seasonal variation: Is it a valid construct? Clinical Psychological Science, 4(5), 825-834. https://doi.org/10.1177/2167702615615867

Treadway, M. T., & Zald, D. H. (2011). Reconsidering anhedonia in depression: Lessons from translational neuroscience. Neuroscience and Biobehavioral Reviews, 35(3), 537-555. https://doi.org/10.1016/j.neubiorev.2010.06.006

Turner, E. H. (2019). Esketamine for treatment-resistant depression: Seven concerns about efficacy and FDA approval. The Lancet Psychiatry, 6(12), 977-979. https://doi.org/10.1016/S2215-0366(19)30394-3

Turner, E. H., Matthews, A. M., Linardatos, E., Tell, R. A., & Rosenthal, R. (2008). Selective publication of antidepressant trials and its influence on apparent efficacy. New England Journal of Medicine, 358(3), 252-260. https://doi.org/10.1056/NEJMsa065779

UK ECT Review Group. (2003). Efficacy and safety of electroconvulsive therapy in depressive disorders: A systematic review and meta-analysis. The Lancet, 361(9360), 799-808. https://doi.org/10.1016/S0140-6736(03)12705-5

Van Orden, K. A., Witte, T. K., Cukrowicz, K. C., Braithwaite, S. R., Selby, E. A., & Joiner, T. E., Jr. (2010). The interpersonal theory of suicide. Psychological Review, 117(2), 575-600. https://doi.org/10.1037/a0018697

Viduani, A., Arenas, D. L., Benetti, S., Wahid, S. S., Kohrt, B. A., & Kieling, C. (2024). Systematic review and meta-synthesis: How is depression experienced by adolescents? A synthesis of the qualitative literature. Journal of the American Academy of Child & Adolescent Psychiatry, 63(10), 970-990. https://doi.org/10.1016/j.jaac.2023.11.013

Wakefield, J. C., Schmitz, M. F., First, M. B., & Horwitz, A. V. (2007). Extending the bereavement exclusion for major depression to other losses: Evidence from the National Comorbidity Survey. Archives of General Psychiatry, 64(4), 433-440. https://doi.org/10.1001/archpsyc.64.4.433

Watkins, E. R. (2008). Constructive and unconstructive repetitive thought. Psychological Bulletin, 134(2), 163-206. https://doi.org/10.1037/0033-2909.134.2.163

Watkins, E. R., & Roberts, H. (2020). Reflecting on rumination: Consequences, causes, mechanisms and treatment of rumination. Behaviour Research and Therapy, 127, 103573. https://doi.org/10.1016/j.brat.2020.103573

Williams, J. M. G., Barnhofer, T., Crane, C., Hermans, D., Raes, F., Watkins, E., & Dalgleish, T. (2007). Autobiographical memory specificity and emotional disorder. Psychological Bulletin, 133(1), 122-148. https://doi.org/10.1037/0033-2909.133.1.122

Williams, J. M. G., Crane, C., Barnhofer, T., Brennan, K., Duggan, D. S., Fennell, M. J. V., Hackmann, A., Krusche, A., Muse, K., Von Rohr, I. R., Shah, D., Crane, R. S., Eames, C., Jones, M., Radford, S., Silverton, S., Sun, Y., Weatherley-Jones, E., Whitaker, C. J., Russell, D., & Russell, I. T. (2014). Mindfulness-based cognitive therapy for preventing relapse in recurrent depression: A randomized dismantling trial. Journal of Consulting and Clinical Psychology, 82(2), 275-286. https://doi.org/10.1037/a0035036

Winter, N. R., Leenings, R., Ernsting, J., Sarink, K., Fisch, L., Emden, D., Blanke, J., Goltermann, J., Opel, N., Barkhau, C., Meinert, S., Dohm, K., Repple, J., Mauritz, M., Gruber, M., Leehr, E. J., Grotegerd, D., Redlich, R., Jansen, A., ... Hahn, T. (2022). Quantifying deviations of brain structure and function in major depressive disorder across neuroimaging modalities. JAMA Psychiatry, 79(9), 879-888. https://doi.org/10.1001/jamapsychiatry.2022.1780

Wisner, K. L., Sit, D. K. Y., McShea, M. C., Rizzo, D. M., Zoretich, R. A., Hughes, C. L., Eng, H. F., Luther, J. F., Wisniewski, S. R., Costantino, M. L., Confer, A. L., Moses-Kolko, E. L., Famy, C. S., & Hanusa, B. H. (2013). Onset timing, thoughts of self-harm, and diagnoses in postpartum women with screen-positive depression findings. JAMA Psychiatry, 70(5), 490-498. https://doi.org/10.1001/jamapsychiatry.2013.87

Woody, C. A., Ferrari, A. J., Siskind, D. J., Whiteford, H. A., & Harris, M. G. (2017). A systematic review and meta-regression of the prevalence and incidence of perinatal depression. Journal of Affective Disorders, 219, 86-92. https://doi.org/10.1016/j.jad.2017.05.003

Yan, C.-G., Chen, X., Li, L., Castellanos, F. X., Bai, T.-J., Bo, Q.-J., Cao, J., Chen, G.-M., Chen, N.-X., Chen, W., ... Zang, Y.-F. (2019). Reduced default mode network functional connectivity in patients with recurrent major depressive disorder. Proceedings of the National Academy of Sciences of the United States of America, 116(18), 9078-9083. https://doi.org/10.1073/pnas.1900390116

Zaninotto, L., Guglielmo, R., Calati, R., Ioime, L., Camardese, G., Janiri, L., Bria, P., & Serretti, A. (2015). Cognitive markers of psychotic unipolar depression: A meta-analytic study. Journal of Affective Disorders, 174, 580-588. https://doi.org/10.1016/j.jad.2014.11.027

Zaninotto, L., Solmi, M., Veronese, N., Guglielmo, R., Ioime, L., Camardese, G., & Serretti, A. (2016). A meta-analysis of cognitive performance in melancholic versus non-melancholic unipolar depression. Journal of Affective Disorders, 201, 15-24. https://doi.org/10.1016/j.jad.2016.04.039

Zimmerman, M., Ellison, W., Young, D., Chelminski, I., & Dalrymple, K. (2015). How many different ways do patients meet the diagnostic criteria for major depressive disorder? Comprehensive Psychiatry, 56, 29-34. https://doi.org/10.1016/j.comppsych.2014.09.007

Zimmerman, M., Martin, J., McGonigal, P., Harris, L., Kerr, S., Balling, C., Kiefer, R., Stanton, K., & Dalrymple, K. (2019). Validity of the DSM-5 anxious distress specifier for major depressive disorder. Depression and Anxiety, 36(1), 31-38. https://doi.org/10.1002/da.22837

Zisook, S., Corruble, E., Duan, N., Iglewicz, A., Karam, E. G., Lanouette, N., Lebowitz, B., Pies, R., Reynolds, C., Seay, K., Shear, M. K., Simon, N., & Young, I. T. (2012). The bereavement exclusion and DSM-5. Depression and Anxiety, 29(5), 425-443. https://doi.org/10.1002/da.21927

Łojko, D., & Rybakowski, J. K. (2017). Atypical depression: Current perspectives. Neuropsychiatric Disease and Treatment, 13, 2447-2456. https://doi.org/10.2147/NDT.S147317