Systematic Review and Meta‐Analysis of the Association Between Subjective and Objective Cognitive Function in Mood Disorders

ABSTRACT

Introduction

The association between subjective and objective cognitive function in mood disorders is hotly debated, particularly in the choice of outcomes to measure pro‐cognitive effects of interventions. We systematically reviewed the evidence regarding this association, including analysis of predicting or moderating factors.

Methods

We searched Ovid MEDLINE, EMBASE, PsycINFO, and Cochrane Library for studies examining subjective and objective cognitive function in mood disorders (from inception to July 2024). Studies using correlational analyses to examine the association between subjective and objective cognitive function were included in the meta‐analysis.

Results

Fifty‐five studies (59 publications) were identified (n = 5798), with 35 studies included in the meta‐analysis. There were weak but statistically significant positive correlations between subjective and objective cognition in all domains except sustained attention, which was at a trend level (p = 0.05). Mood state or mood disorder diagnosis did not impact these correlations. No subjective measure was more highly associated with objective cognitive function than others. Our narrative review of the remaining 20 studies (that did not examine subjective‐objective correlations) demonstrated a shift towards calculation of a discrepancy or sensitivity score, rather than simply correlating measures.

Conclusion

This is the first systematic review and meta‐analysis examining the association between subjective and objective cognition in mood disorders. Our results support the general consensus that there is not a strong relationship. This may relate to the limited ecological validity of objective cognitive tests and highlights the need for assessment of both subjective and objective cognitive function to adequately capture patients' cognitive status.

1. Introduction

Cognitive impairment is common during symptomatic and remitted phases in major depressive disorder (MDD) [1, 2, 3, 4] and bipolar disorder (BD) [5, 6] with prevalence rates as high as 20%–40% [7, 8] and 34%– 49% [9, 10] respectively. Objectively measured (performance‐based) impairment occurs across cognitive domains, including sustained attention, psychomotor speed, learning and memory, and executive function [3, 11]. These impairments contribute to problems in functioning and quality of life, as well as increased risk of relapse [12, 13, 14, 15, 16]. Subjective (self‐reported) cognitive difficulties are also common, being reported by 63%–85% of people with mood disorders [17, 18], and may relate more strongly to functioning [17, 19]. Indeed, the perception of cognitive problems is one of the most common symptoms in MDD [20]. Therefore, both subjective and objective cognitive impairment may be targets for treatment aimed at improving functional recovery in people with mood disorders.

Treatment trials aimed at improving cognitive function in mood disorders are limited by a number of methodological issues. One key issue is the current incomplete understanding of the nature and pattern of cognitive impairment. Determining the level of cognitive impairment is compounded by evidence demonstrating discrepancies between subjectively experienced and objectively measured cognitive impairment in BD and MDD [19, 21]. Patients with the greatest cognitive complaints regarding cognition may thus not display the most objective impairment, and vice versa.

Several factors may be implicated in the seemingly poor association between subjective and objective cognitive function. First, it may be that different concepts are being addressed when asking about cognition in daily life compared with performance on objective cognitive tests, and this may depend on the domain being measured. For instance, memory and attention difficulties may be more readily identifiable and understood by patients than executive functioning or psychomotor speed problems. There may, thus, be some domains of cognition where subjective and objective cognitive function are more highly related [21]. Second, objective cognitive function is tested under controlled, quiet conditions using clear instructions, allowing for a detailed examination of specific domains. This differs from cognitive challenges experienced by people with mood disorders in their daily life, where multiple domains of cognition are required simultaneously, lots of distractions occur and no clear instructions are provided. Third, other factors may be important, including severity of depressive symptoms [22, 23] and mood state including mania and mixed states [24], which may moderate objective and subjective cognition differentially. Finally, it is possible that some patients with higher‐than‐normal premorbid cognitive function experience a decline after illness onset but that this cannot be captured with cognitive tests that merely involve comparison with a normative sample and thus show cognitive performance in the normal range [25].

While a poor correlation between subjective and objective cognitive function is often cited and studied, there has been no systematic review of the evidence with regard to this association in mood disorders. Better understanding the association between these aspects of cognitive function may have clinical implications with regard to how clinicians assess the level of impairment patients are experiencing and research implications in terms of more accurate measurement of cognitive function. There may also be implications for treatment, including improved identification of patients who are at risk of or experiencing different types of cognitive impairment and targeted treatments aimed at remediating these. Therefore, our aim was to systematically review the evidence regarding the association between subjective and objective cognitive function in mood disorders. We also aimed to identify factors that consistently predicted or moderated the association.

2. Materials and Methods

A systematic review of the literature, following the Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA) guidelines [26], was conducted to identify studies examining the association between subjective and objective cognitive functioning in mood disorders. The study protocol was developed prior to searching electronic databases and was registered in the International Prospective Register of Systematic Reviews (PROSPERO, CRD42021265298).

2.1. Search Strategy

Ovid MEDLINE, EMBASE, PsycINFO, and Cochrane Library were searched from inception through August 2023. There was no limit on date of publication. The databases were searched using the following terms:

1. ‘bipolar disorder’ or ‘bipolar I disorder’ or ‘bipolar II disorder’ or ‘major depressive disorder’ or ‘depressive disorder, major’ or ‘major depression’ or ‘mood disorders’ or ‘affective disorders’ AND

2. ‘cognition’ or ‘cognitive function*’ or ‘cognitive dysfunction’ or ‘neurocognition’ or ‘objective cognitive function*’ or ‘cognition assessment’ or ‘cognitive impairment’ or ‘cognitive ability’ AND

3. ‘subjective cognitive function*’ or ‘cognitive complaint*’ or ‘self‐report’ or ‘patient reported outcome measures’ or ‘self‐evaluation’

Titles and abstracts of articles identified were thoroughly assessed by two authors (KE and KD) to confirm or reject their inclusion based on prespecified selection criteria. When a decision was not possible from reading the abstract, the full text was reviewed and contrasted with inclusion and exclusion criteria. Full texts were screened by two authors (KE and KD) independently for inclusion. KE and KD discussed any discrepancies and if consensus was unable to be reached, full texts were reviewed by RP. Manual searching was conducted using reference lists in studies identified by the initial search. The search was re‐run in July 2024, prior to final analyses to identify further papers.

2.2. Study Selection Criteria

Studies were considered for the review if they met the following inclusion criteria:

1. measurement of subjective and objective cognitive function;

2. examination of the association between subjective and objective cognitive function;

3. primary diagnosis of MDD or BD (using Diagnostic and Statistical Manual (DSM) or International Statistical Classification (ICD) criteria);

4. adult population (aged 18 years and over);

5. publication in peer‐reviewed journals;

6. published in English.

Exclusion criteria were:

1. studies where outcomes of patients with mood disorders were not able to be separated from those with other psychiatric disorders;

2. primary diagnosis of comorbid major medical condition;

3. diagnosis of dementia or mild cognitive impairment;

4. systematic reviews, books, commentaries, conference abstracts or dissertations.

2.3. Data Extraction

For each study, the following data were extracted: sample characteristics (sample size, diagnosis, mean age, baseline mood symptoms), study design, cognitive measures, outcomes. Data were extracted by KE and reviewed by the other authors. See Figure 1 for the PRISMA flow diagram of the review process.

FIGURE 1.

PRISMA flow diagram.

2.4. Risk of Bias Assessment

The Joanna Briggs checklist for cross sectional studies was used as a formal risk of bias tool [27]. See Table 1 for risk of bias ratings for each study. Quality assessment was performed by two authors (K.E. and R.P.). Disagreements regarding quality scores for each study were resolved by discussion. Studies were weighted by quality when synthesising the evidence for the narrative review.

TABLE 1.

Risk of bias summary.

Note: 1—Were the criteria for inclusion in the sample clearly defined? 2—Were the study subjects and setting described in detail? 3—Was the exposure measured in a valid and reliable way? 4—Were objective, standard criteria used for measurement of the condition? 5—Were confounding factors identified? 6—Were strategies to deal with confounding factors stated? 7—Were the outcomes measured in a valid and reliable way? 8—Was appropriate statistical analysis used?

2.5. Meta‐Analyses

Studies using correlational analyses to examine the association between subjective and objective cognitive function were included in the meta‐analysis. For studies with incomplete data presentation, authors were contacted to request the relevant results. Objective cognitive tests were grouped into five cognitive domains: processing speed, sustained attention, verbal learning and memory, visual learning and memory, working memory/executive function (see Table S1). Where studies used more than one objective test in a domain, the most commonly used test across studies was included in the meta‐analyses. Both global (e.g., Cognitive Complaints in Bipolar Disorder Rating Assessment [COBRA]) and domain‐specific (e.g., Behaviour Rating Inventory of Executive Function [BRIEF‐A]) subjective cognitive measures were extracted, with global measures used preferentially for the primary meta‐analysis. Correlations were adjusted so that a positive correlation reflected concordance between objective and subjective cognitive function (i.e., poorer functioning in both). Sub‐analyses were undertaken investigating the impact of mood disorder diagnosis (MDD vs. BD), mood state (remitted vs. non‐remitted), and subjective cognitive measure (global vs. domain‐specific measures). The association between global objective cognitive function and different subjective cognitive measures was examined. Where studies had used a composite objective measure that was not in keeping with our pre‐specified domains, sensitivity analyses were conducted with and without the study included. Several studies that would have been eligible for meta‐analysis provided data on significant correlations only and were therefore excluded.

3. Results

3.1. Study Selection and Characteristics

The systematic search, together with the additional hand‐search, identified 1524 articles (after removal of duplicates) that were included for title/abstract screening (primary screening). Of these, 134 were evaluated for eligibility via a full‐text reading (secondary screening). This resulted in the inclusion of 59 articles that met the inclusion criteria (see Figure 1). Fifty‐five studies (59 publications) meeting inclusion criteria were identified, with 5798 participants (see Figure 1 for PRISMA flow diagram). Thirty‐five studies were included in the meta‐analysis. Twenty studies did not examine the correlation between subjective and objective cognition and were therefore excluded from the meta‐analysis. However, they are included in the narrative review below. See Tables 2 and 3 for characteristics of included studies.

TABLE 2.

Characteristics of included studies in meta‐analysis (n = 35).

Author, year	N	Age	Mood state	Diagnosis	Objective cognitive measures	Subjective cognitive measures
Aydemir et al., 2017 [28]	50	18–65 years M = 37.5 (11.4)	Non‐remitted	MDD	DSST	BC‐CCI
Beblo et al., 2017 [29]	20	M = 41.4 (12.5)	Non‐remitted	MDD	CFT Supermarket word list recall ELM	FLei
Bernhardt et al., 2021 [30]	103	18–65 years M = 43.37 (12.12)	Non‐remitted	MDD	WAFA WAFG FGT TMT‐B Go‐No‐Go TOL	FLei
Bickford et al., 2018 [31]	247	> 65 years M = 71.3 (6.0)	Non‐remitted	MDD	TMT‐B SCWT	PDQ (EF items)
Burdick et al., 2005 [32]	37	M = 46.2 (14.1)	Non‐remitted	BD	CVLT Digit span Digit symbol SCWT TMT‐A & B	CFQ
Dalkner et al., 2023 [33]	113	M = 38.45 (11.69)	Non‐remitted	BD	LNS TMT‐A BACS symbol coding LNS Animal naming test HVLT	CAI
Demant et al., 2015 [19]	77	M = 37.4 (10.5)	Non‐remitted	BD	RAVLT RVP RBANS coding and digit span TMT‐B LNS Verbal and semantic fluency	CPFQ
Faurholt‐Jepsen et al., 2020 [34]	117	M = 30.9 (9.9)	Remitted	BD	SCIP	CPFQ
Fava et al., 2018 [35]	602	18–65 years M = 44.2–45.7 (11.5–12.2)	Non‐remitted	MDD	DSST TMT‐A TMT‐B SCWT GMLT Detection task Identification task CRT One‐Back task	CPFQ
Ingulfsvann Hagen et al., 2021 [36]	62	18–60 years M = 42.1 (8.5)	Non‐remitted	MDD	WCST TMT Colour‐Word Interference Test CPT Digit Span B LNS	BRIEF‐A
Harvey et al., 2015 [37]	30	M = 44.3 (12.8)	Non‐remitted	BD	MCCB WCST	CAI
Jensen et al., 2015 [17]	84	M = 36 (10)	Remitted	BD	SCIP RAVLT LNS Digit span forward Verbal fluency TMT‐A	COBRA
Keilp et al., 2018 [38]	262	18–72 years M = 38.1 (12.0)	Non‐remitted	MDD	CRT Digit Symbol CPT‐IP SCWT BSRT A, Not B Logical Reasoning test Letter Fluency Category Fluency WCST Go‐No‐Go	CFQ
Lima et al., 2018 [39]	78	18–70 years M = 49.60 (12.88)	Remitted	BD	F‐A‐S TMT‐A LNS HVLT‐R SCWT TMT‐B CPT‐IP	COBRA
Lin et al., 2019 [40]	140	18–60 years M = 34.08–34.94 (8.16–9.85)	Non‐remitted	BD	Category fluency SCWT TMT‐B Digit Span F & B Digit Symbol Coding TMT‐A WMS (Visual Reproduction and Visual Recognition)	COBRA
Matcham et al., 2023 [41]	448	M = 46.4 (15.3)	Non‐remitted	MDD	THINC‐it	PDQ‐5
Mendes et al., 2021 [42]	47	M = 59.3 (8.7)	Non‐remitted	MDD	MMSE WMS	SMCS
Miskowiak et al., 2024 [10]	136	Median = 29 (IQR = 24, 36)	Non‐remitted	BD	SCIP	COBRA
Miskowiak et al., 2021 [43]	30	18–60 years M = 31.7 (9.0)	Remitted	BD	SCIP	COBRA
Mohn and Rund, 2016 [44]	33	18–70 years M = 46.5 (10.6)	Non‐remitted	MDD/BD	MCCB	EMQ
Naismith et al., 2007 [45]	21	M = 53.9 (11.0)	Non‐remitted	MDD	TMT‐A WMS logical memory RAVLT TMT‐B SCWT TOL	Likert scale of cognitive function compared with pre‐morbid cognitive function
Ott et al., 2016 [18]	38	M = 41.8 (12.1)	Non‐remitted	MDD	SCIP RAVLT Digit Span Verbal fluency DLST TMT‐A	COBRA
Ott et al., 2019 [46]	58	18–65 years M = 37 (11)	Remitted	BD	RAVLT LNS Digit Span Verbal Fluency TMT‐B SWM OTS RVP TMT‐A RBANS coding	COBRA
Potvin et al., 2016 [47]	40	M = 51.23 (11.43)	Non‐remitted	MDD	SCIP	EDEC
Schwert et al., 2018 [48]	102	18–66 years M = 42.62 (12.97)	Non‐remitted	MDD	COGBAT: WAFA WAFG FGT TMT‐B Go‐No‐Go TOL	FLei
Serra‐Blasco et al., 2019 [49]	229	18–65 years M = 51.09–53.99 (6.64–11.68)	Non‐remitted	MDD	TMT‐A Digit Span F & B Spatial Span F & B RAVLT	PDQ – attention and memory subscales
Shi et al., 2017 [50]	129	18–65 years M = 40.61 (14.22)	Non‐remitted	MDD	DSST	PDQ
Srisurapanont et al., 2017 [51]	57	21–65 years M = 45.1 (12.8)	Non‐remitted	MDD	WMS (Face I and II) Digit Span Matrix Reasoning	PDQ
Tourjman et al., 2019 [52]	40	M = 50.8 (1.8)	Non‐remitted	MDD	SCIP	PDQ
Toyoshima et al., 2017 [53]	41	18–64 years M = 43.34 (10.51)	Remitted	BD	WCST WFT CPT TMT AVLT SCWT	COBRA
van Camp et al., 2019 [54]	29	M = 39.43 (10.69)	Non‐remitted	BD	ISDB‐BANC: TMT A Symbol Coding Category fluency (animals) CPT‐IP Spatial Span LNS HVLT‐R Mazes TMT B SCWT	Likert scale of general cognitive functioning
van der Werf‐Eldering et al., 2011 [55]	108	18–65 years M = 45.8 (10.7)	Non‐remitted	BD	Reaction time test SCWT CPT CVLT PRM SWM	CFQ
Vicent‐Gil et al., 2023 [56]	324 Non‐remitted n = 208 Remitted n = 117	Non‐remitted M = 52.55 (7.69) Remitted M = 51.56 (9.73)	Non‐remitted	MDD	Digit Span F TMT‐A & B RAVLT Phonemic verbal fluency WCST DSST	FAST cognitive domain
Wang et al., 2019 [57]	598	18–65 years M = 36.5 (12)	Non‐remitted	MDD	DSST	PDQ
Xiao et al., 2016 [58]	100	M = 26.70 (9.18)	Remitted	BD	MoCA	COBRA

Abbreviations: BC‐CCI, British Columbia Cognitive Complaints Inventory; BRIEF‐A, Behaviour Rating Inventory of Executive Function – Adult; BSRT, Buschke Selective Reminding Test; CAI, Cognitive Assessment Interview; CFQ, Cognitive Failures Questionnaire; CFT, Complex Figure Test; COBRA, Cognitive Complaints in Bipolar Disorder Rating Assessment; CPFQ, Cognitive and Physical Functioning Questionnaire; CPT, Continuous Performance Test; CRT, Choice Reaction Time; CVLT, California Verbal Learning Test; DLST, Digit Letter Substitution Test; DSST, Digit symbol substitution test; EDEC, Échelle d'auto‐évaluation cognitive; ELM, Everyday Life Memory; EMQ, Everyday Memory Questionnaire; F‐A‐S, Phonemic Verbal Fluency; FAST, Functioning Assessment Short Test; FGT, Figural Memory Test, questionnaire for cognitive complaints, subscale executive functions; FLei, Fragebogen zur geistigen Leistungsfaehigkeit; GMLT, Groton Maze Learning Test; HVLT, Hopkins Verbal Learning Test; LNS, Letter Number Sequencing; MCCB, MATRICS Consensus Cognitive Battery; MMSE, Mini Mental State Examination; MoCA, Montreal Cognitive Assessment; OTS, One Touch Stockings of Cambridge; PDQ, Perceived Deficits Questionnaire; PRM, Pattern Recognition Memory; RAVLT, Rey Auditory Verbal Learning Test; RBANS, The Repeatable Battery for the Assessment of Neuropsychological Status; RVP, Rapid Visual Information Processing; SCIP, Screen for Cognitive Impairment in Psychiatry, Self‐Perception of Cognitive Performance in everyday life and test settings; SCWT, Stroop Colour and Word Test; SKAT, Selbstwahrnehmung kognitiver Leistungen im Alltag und Testsituation; SMCS, Subjective Memory Complaints Scale; SWM, Spatial Working Memory; TMT, Trail Making Test; TOL, Tower of London; WAFA, WAF test battery alertness; WAFG, divided attention; WCST, Wisconsin Card Sorting Test; WFT, Word Fluency Test; WMS, Wechsler Memory Scale.

TABLE 3.

Characteristics of studies in narrative review (n = 20).

Author, year	N	Age	Mood state	Diagnosis	Objective cognitive measures	Subjective cognitive measures	Main findings
Beblo et al., 2023 [59]	58	M = 41.9 (10.8)	Non‐remitted	MDD	SCIP	FLei‐EF	No significant association
Beblo et al., 2020 [60]	22	M = 42.5 (12.4)	Non‐remitted	MDD	WMS logical memory	FLei	No significant association
Beblo et al., 2010 [61]	30	M = 37.7 (12.9)	Non‐remitted	MDD	AVLT	FEDA EMQ	No significant association
Bonnin et al., 2024 [62]	83	M = 43.9 (10.4)	Remitted	BD	Digit symbol coding Symbol search TMT‐A LNS WCST SCWT TMT‐B F‐A‐S COWAT animals CVLT	COBRA	Global sensitivity associated with higher depressive episodes, lower number of previous hospitalisations and FAST
Brakemeier et al., 2011 [63]	90	M = 44.9–56.8 (11.5–18.8)	Non‐remitted	MDD	CUAMI MMSE ROCF BSRT	CFQ SMCQ GSE‐My	Association between GSE‐My and CUAMI, BSRT. Trend association with ROCF. No association with MMSE. No association between CFQ or SMCQ and objective measures
Feehan et al., 1991 [64]	10	> 65 years M = 74.6 (7.2)	Non‐remitted	MDD	WMS: logical memory and visual reproduction VOT LCT MMSE	CFQ	No significant association
Jensen et al., 2016 [65]	193	18–65 years M = 36 (10)	Remitted	BD	TMT‐A RAVLT LNS Digit Span F Verbal fluency	CPFQ	Selectively and globally cognitively impaired patients had more subjective cognitive impairment compared with those without objective impairment
Lahr et al., 2007 [66]	15	18–55 years M = 38.1	Non‐remitted	MDD	TMT‐A + B Test d2 TAP AVLT DST BST Lexical fluency Semantic fluency FPT	QEAD QME	No significant association
Lima et al., 2019 [67]	73	18–70 years M = 49.29 (12.66)	Remitted	BD	HVLT‐R SCWT TMT‐A TMT‐B F‐A‐S CPT‐IP LNS	COBRA	Similar pattern of subjective cognition in intact, selective and global objective cognitive impairment groups
MacQueen et al., 2001 [68]	28	M = 42.2 (12.3)	Remitted	BD	VBM	CFQ	No significant association
Martinez‐Aran et al., 2005 [69]	60	M = 38.5 (9.9)	Remitted	BD	WCST SCWT COWAT Digit span TMT CVLT WMS (logical memory, visual reproduction)	Clinical interview (≥ 1 ‘significant subjective complaint’)	Patients with subjective impairment had poorer objective performance than patients without subjective impairment in most tests, except EF
Miskowiak et al., 2016 [70]	109	28–65 years M = 36 (10)	Remitted	BD	RAVLT Digit span F LNS Verbal fluency TMT‐A SCIP	COBRA	Greater stoicism in attention and processing speed compared with verbal learning and memory and WM/EF
Mowla et al., 2008 [71]	64	M = 32.3 (9.6)	Non‐remitted	MDD	WMS: General information Orientation Logical memory Verbal paired associates Mental control Digit span Family pictures	Global evaluation of memory (4‐point Likert)	No significant difference in WMS total score between groups (with or without subjective impairment)
Otto et al., 1994 [72]	156	18–65 years M = 38.9 (9.8)	Non‐remitted	MDD	CVLT	CFQ	No significant association
Petersen et al., 2019 [21]	157	M = 38–42	Non‐remitted	MDD/BD	RAVLT Digit Span F LNS COWAT TMT‐A SCIP Motor Screening Spatial Span SWM CRT CVC	COBRA Items from BDI‐II and SCL‐90	Discrepancy between subjective and objective cognitive function across all domains
Quinlivan et al., 2023 [73]	26	M = 44.91 (11.98)	Remitted	BD	SCWT Go‐No‐Go TAP alertness Divided attention Digit span F and B VLMT MWT‐B Fluency (S, animals)	QEAD	No significant association
Rnic et al., 2021 [74]	124	18–60 years M = 34.93 (12.43)	Non‐remitted	MDD	CNS‐VS	DID cognition items	Accurate self‐appraisal of cognition at baseline. Overestimation of cognitive function post‐treatment
Schouws et al., 2012 [75]	101	> 60 years M = 67.8–70.7 (6.9–7.4)	Non‐remitted	BD	Digit Span TMT‐A 10 Words Test Figure Copying Clock Drawing TMT‐B SCWT Mazes Rule Shift Cards COWAT Animal and Occupation Naming	CFQ	Patients with less subjective cognitive impairment had poorer performance in attention and EF tests than patients with more subjective cognitive impairment
Svendsen et al., 2012 [76]	30	18–65 years M = 34–41 (9.1–14.2)	Non‐remitted	MDD/BD	SCIP	COBRA	No significant association
Veeh et al., 2017 [77]	26	18–60 years M = 36.3–42.3 (12.2–12.3)	Remitted	BD	SCWT TAP – divided attention, working memory CVLT TOL	Flei	No significant association

Abbreviations: BDI‐II, Beck Depression Inventory II; BSRT, Buschke Selective Reminding Test; BST, Block Suppression Test; CFQ, Cognitive Failures Questionnaire; CNS‐VS, Central Nervous System Vital Signs; COBRA, Cognitive Complaints in Bipolar Disorder; COWAT, Controlled Oral Word Association Test; CPFQ, Cognitive and Physical Functioning Questionnaire; CPT, Continuous Performance Test; CRT, Choice Reaction Time; CUAMI, Columbia Autobiographical Memory Interview; CVC, Consonant Vowel Consonant; CVLT, California Verbal Learning Test; DID, Depression Inventory Development; DST, Digit Suppression Test; EMQ, Everyday Memory Questionnaire; F‐A‐S, Phonemic Verbal Fluency; FAST, Functioning Assessment Short Test; FEDA, Questionnaire of Experienced Attention (Fragebogen zur erlebten Aufmerksamkeit); FLei, Questionnaire for cognitive complaints subscale executive functions (Fragebogen zur geistigen Leistungsfaehigkeit); FPT, Five Point Test; GSE‐My, Global Self Evaluation of Memory; HVLT, Hopkins Verbal Learning Test; LCT, Letter Cancellation Task; LNS, Letter Number Sequencing; MMSE, Mini Mental State Examination; MWT‐B, Mehrfachwahl‐Wortschatz‐Intelligenztest; QEAD, Questionnaire of Experienced Attention Deficits; QME, Questionnaire of Memory Efficiency; RAVLT, Rey Auditory Verbal Learning Test; ROCF, Rey–Osterrieth complex figure; SCIP, Screen for Cognitive Impairment in Psychiatry; SCL‐90, Symptom Checklist‐90; SCWT, Stroop Colour and Word Test; SMCQ, Squire Memory Complaints Questionnaire; TAP, Test of Attentional Performance; TMT, Trail Making Test; TOL, Tower of London; VBM, Visual backward masking; VLMT, Verbal Learning and Memory Test; VOT, Visual Organisation Test; WCST, Wisconsin Card Sorting Test; WMS, Wechsler Memory Scale.

Roughly equal numbers of studies included participants with MDD (n = 28) and BD (n = 24), and three studies included mixed samples with MDD and BD. Most participants had MDD (n = 4071, 70.2%). MDD studies were generally larger. While the majority of studies were in adult samples, four were in older adults without dementia [31, 42, 64, 75]. Two studies [65, 70] analysed pooled data from other included studies [17, 18, 19]. Five studies selected participants based on cognitive impairment, with three selecting participants with subjective impairment [19, 35, 36] and two with objective impairment [46, 77].

The majority of studies were cross‐sectional (k = 49), with only six studies (11.8%) involving longitudinal cognitive assessments. These included a Cognitive Behavioural Therapy (CBT) intervention [30], a small naturalistic treatment study [66], a longitudinal Electroconvulsive Therapy (ECT) study [44], an open‐label clinical trial of escitalopram [74], a naturalistic study of a Cognitive Remediation (CR) intervention [77], and one observational study [41].

There was significant heterogeneity across studies in terms of clinical characteristics and choice of measures. Baseline mood state varied, with studies including depressed (k = 35), partially remitted (k = 16), remitted (k = 25), manic or hypomanic (k = 2), and mixed mood (k = 1) patients (see Tables 2 and 3). Fifteen studies included participants in a variety of mood states. A wide variety of measures were used for both subjective and objective cognitive function; 22 different subjective cognitive measures were used, and 85 different objective cognitive tests were used. Regarding subjective measures, the Cognitive Complaints in Bipolar Disorder Rating Assessment (COBRA) (k = 14) and Cognitive Failures Questionnaire (CFQ) (k = 10) were the most utilised (43.6% of studies). For objective cognitive assessments, some studies used standardised cognitive batteries (e.g., the Screen for Cognitive Impairment in Psychiatry; SCIP: k = 11), while the majority (k = 34) used their own customised battery of tests including 37 tests. Most of these tests were used in multiple studies (see Table S1). Seven studies used only one objective cognitive measure, probing processing speed (k = 3), verbal learning and memory (k = 3), and visual learning and memory (k = 1). Finally, four studies used dementia screening tests (Montreal Cognitive Assessment [MoCA], Mini‐Mental State Examination [MMSE]).

The risk of bias evaluations of included studies are presented in Table 1. Sixteen studies were rated as ‘low risk’, while 36 were rated as ‘some concerns’ and three as ‘high risk’. Where methodological issues were present, this was generally due to studies that did not measure or adjust for demographic or clinical confounders (e.g., IQ, medications, and mood symptoms). We rated these studies as ‘some concerns’. The recruitment strategy was unclear in several studies, which were therefore rated as ‘high risk’.

3.2. Meta‐Analysis

Thirty‐five (63.6%) of the identified studies analysed the correlation between subjective and objective cognitive function and were thus eligible to be included in the meta‐analysis. Three studies did not report raw correlations and authors were contacted, with one study providing data. The meta‐analysis included a total of 4630 participants. Overall, there was significant heterogeneity in the results of correlational analyses examining processing speed (I ² = 54.3%, p = 0.003), sustained attention (I ² = 70.2%, p = 0.003), and verbal learning and memory domains (I ² = 45.96%, p = 0.03). There was no significant heterogeneity in the global, visual learning and memory, or executive functioning and working memory domains (p all > 0.05).

There were weak but statistically significant correlations between subjective and objective cognition in the following domains: processing speed (n = 3331; r = 0.17, p < 0.001); verbal learning and memory (n = 1472; r = 0.16, p < 0.001); visual learning and memory (n = 398; r = 0.12, p = 0.02); executive functioning and working memory (n = 3037; r = 0.11, p < 0.001); and global cognition (n = 2213; r = 0.18, p < 0.001) (Figure 2). The sustained attention domain was weakly correlated at a trend level (n = 653; r = 0.16, p = 0.05). Confidence intervals were overlapping between domains (Figure 3).

FIGURE 2.

Relationship between subjective and objective cognition within domains.

FIGURE 3.

Relationship between subjective and objective cognition across domains.

Mood state did not have a consistent effect on the correlation between subjective and objective cognitive function (Figure 4). Correlations between subjective and objective cognitive function between MDD and BD samples were overlapping (Figure S1). Use of domain‐specific subjective cognition measures did not lead to stronger correlations with objective cognition; however, sample sizes were small (Figure S2).

FIGURE 4.

Effect of mood state on relationship between subjective and objective cognition.

We assessed the correlation between different global subjective cognitive measures and global objective cognitive function. This analysis included 16 studies (n = 2115) and five subjective cognitive measures. There were weak statistically significant correlations using four of the scales: the COBRA (n = 486, r = 0.26, p < 0.001), the Cognitive and Physical Functioning Questionnaire (CPFQ) (n = 796, r = 0.14, p < 0.001), the Perceived Deficits Questionnaire (PDQ) (n = 545, r = 0.16, p = 0.009), and the Cognitive Assessment Interview (CAI) (n = 143, r = 0.22, p = 0.009) (Figure S3). The analysis for the Cognitive Failures Questionnaire (CFQ) was not statistically significant (r = 0.08, p = 0.33); however, this included only 145 patients.

3.3. Interim Summary

Our meta‐analysis demonstrates a weak, statistically significant association between subjective and objective cognitive function across global measures and several cognitive domains. While sustained attention was only correlated at a trend level, this analysis had a small sample size. There was no impact of mood state or mood disorder diagnosis on the correlation. All subjective cognitive measures demonstrated a similar strength of correlation between subjective and objective cognitive function.

3.4. Narrative Review

Twenty studies could not be included in the meta‐analysis due to lack of correlational analysis (k = 11), correlation result not presented (k = 8), or analysis between non‐corresponding domains of subjective and objective cognitive function (e.g., correlating subjective executive function with objective global cognition) (k = 1). Half of these studies (k = 10) found there was no significant relationship between subjective and objective cognitive function [59, 60, 61, 64, 66, 68, 72, 73, 76, 77]. Four studies examined discrepancies between subjective and objective cognitive function [21, 62, 70, 74], discussed further below. Five studies examined between‐group differences. Two of these studies split their samples into cognitive subgroups (e.g., no objective impairment, impairment in some domains, global impairment) [65, 67], and three studies analysed their samples based on subjective impairment (with or without) [69, 71, 75]. There were no consistent findings across these studies.

One study found a strong association between subjective memory measured with the Global Self Evaluation of memory (GSE‐My) and an objective measure of autobiographical memory, the Columbia University Autobiographical Memory Interview (CUAMI) (F = 15.42, p = 0.0002), and the Buschke Selective Remining Test (BSRT) (F = 4.73, p = 0.03), post‐ECT [63]. There was no association between other subjective measures and any objective measures [63].

3.5. Interim Summary

Half of the studies in the narrative review found no significant association between subjective and objective cognitive function. Several studies examined between‐group differences when patients were classified as having or not having significant cognitive complaints—but there were no consistent findings.

3.6. Examination of Factors Influencing the Subjective‐Objective Association

Twenty studies included in the systematic review examined factors influencing the association between subjective and objective cognitive function. A variety of methods were used (see Table S2 for detail). Seven studies adjusted for mood symptoms with no consistent findings [19, 34, 40, 53, 54, 55]. Four studies examined the effect of age with mixed results [19, 34, 65, 75]. Five studies examined the impact of IQ, with three demonstrating an association between discrepancy and IQ [36, 49, 70] and two with mixed results on regression analysis [63, 75].

3.7. Discrepancy Between Subjective and Objective Cognitive Function

Eleven studies examined the discrepancy between objective and subjective cognitive function (k = 6 also included in the meta‐analysis). Eight of these studies subtracted objective from subjective cognitive scores (or vice versa) using z‐ or t‐scores [30, 36, 42, 48, 49, 51, 54, 74]. A further three studies assessed whether an individual was “sensitive” or “stoic” to cognitive impairment (i.e., disproportionately worse subjective cognition than objective cognition and vice versa), based on rank ordering [21, 62, 70].

Five of these 11 studies found that patients reported higher levels of subjective cognitive difficulties than objective impairment [30, 36, 42, 48]. However, two studies of MDD patients in a depressive episode found mean discrepancy scores of zero [51, 74], suggesting roughly equal degrees of over‐ and underestimation of performance within the groups. Where included, healthy control participants were either accurate [42] or overestimated [48, 54] their objective performance. See Table 4 for factors contributing to the discrepancy between subjective and objective cognitive function.

TABLE 4.

Factors contributing to discrepancy between subjective and objective cognitive function (n = 11).

Study	Sample details	Underestimation of objective performance/sensitivity	Overestimation of objective performance/stoicism	Relationship with confounders
Bernhardt et al., 2021 [30]	N = 103 MDD Non‐remitted	Baseline, prior to CBT (reduced discrepancy following treatment)
Bonnin et al., 2024 [62]	N = 83 BD Remitted	Depressive episodes FAST Years of education Higher IQ	Number of hospitalisations	Positive correlation between global sensitivity and depressive symptoms and years of education
Ingulfsvann Hagen et al., 2021 [36]	N = 62 MDD Non‐remitted	Higher IQ		No correlation between discrepancy and depressive symptoms
Mendes et al., 2021 [42]	N = 47 MDD Non‐remitted			No correlation between discrepancy and depressive symptoms, age or years of education Positive correlation between discrepancy and objective verbal memory
Miskowiak et al., 2016 [70]	N = 109 BD Remitted	Depressive and manic symptoms Younger age BD II Number of hospitalisations Lower IQ Male gender	Attention and processing speed domain
Petersen et al., 2019 [21]	N = 157 MDD/BD Non‐remitted	Depressive symptoms Younger age Longer illness duration Female Attention and processing speed domain	Greater number of hospitalisations
Rnic et al., 2021 [74]	N = 124 MDD Non‐remitted		Increased stoicism over course of treatment with escitalopram (from concordance)	Negative correlation between discrepancy and depressive symptoms. No correlation with age or years of education
Schwert et al., 2018 [48]	N = 102 MDD Non‐remitted	MDD participants	Healthy control participants
Serra‐Blasco et al., 2019 [49]	N = 229 MDD Non‐remitted	Depressed subgroup Higher IQ	Mildly depressed/partially remitted subgroup	Negative correlation between discrepancy and depressive symptoms, and objective executive functioning IQ not correlated
Srisurapanont et al., 2017 [51]	N = 57 MDD Non‐remitted	Depressive symptoms		Negative correlation between age and discrepancy
Van Camp et al., 2019 [54]	N = 29 BD Non‐remitted	Depressed subgroup	Mildly depressed/partially remitted subgroup Healthy control participants

3.8. Interventional and Longitudinal Studies

Two of the six identified longitudinal studies found no significant correlation between subjective and objective cognitive function at any point [30, 77]. Two studies found either stronger correlations [66] or a reduced discrepancy [30] between subjective and objective cognitive function following treatment for depression. One study found that while subjective and objective cognitive functioning were not related at baseline, some aspects were correlated at 6 weeks post‐ECT; however, this was not sustained at 6 months [44, 78, 79]. Another found concordance between subjective and objective cognition at baseline, but overestimation of objective cognitive function (i.e., greater objective than subjective cognitive impairment) following treatment with escitalopram [74].

3.9. Interim Summary

A relatively small number of included studies (21 of 55, 38%) adjusted for the effects of potential confounders. Very few studies adjusted for the effect of IQ (k = 5); however, four of these found that it impacted the subjective‐objective relationship. Calculation of a discrepancy or sensitivity score, rather than simply correlating measures, allowed for the degree and direction of the discrepancy between subjective and objective cognitive function to be established. However, there were no consistent findings regarding the impact of age, gender, or years of education on discrepancy scores. The impact of mood symptoms may depend on the mood state of included patients.

4. Discussion

This systematic review and meta‐analysis identified 55 studies investigating the association between subjective and objective cognitive function in mood disorder samples. Of these, 35 studies including 4630 participants provided data which enabled meta‐analysis of correlations between global and domain‐specific subjective and objective cognition measures. There were weak but statistically significant correlations between subjective cognitive function and global objective cognition and within all objective domains except sustained attention, with similar correlation strengths across domains. This is novel given the large number of individual studies reporting no association between subjective and objective cognitive function. We identified no significant impact of remission status on these associations, another novel finding given the general consensus that mood state impacts on the relationship between subjective and objective cognitive function [30, 49, 70].

There are several potential explanations for the weak association between subjective and objective cognitive function; the way in which cognitive function is measured, clinical characteristics, and methodological limitations of the current evidence base.

4.1. Measurement of Cognitive Function

Measurement of objective cognitive function may have low ecological validity [66]. Performance is measured on a circumscribed task in a structured environment, with minimal distractions. This may have little relation to daily life where multiple cognitive processes are involved simultaneously and completion of tasks may be influenced more strongly by external factors (e.g., distraction, emotional responses). Subjective cognitive complaints may better reflect the experience of cognitive functioning under less controlled situations [48, 66]. Objective cognitive testing may not be sensitive enough to detect subtle cognitive deficits experienced by patients, or the testing environment may not adequately mimic the cognitive functions required to function in day‐to‐day life. Additionally, the Hawthorne effect (i.e., modification of behaviour in response to awareness of being observed) may contribute to better objective performance under test conditions, even in a paradigm that mimics everyday life [29].

The types of cognitive dysfunction that are most noticeable to patients may not be commonly measured by objective cognitive testing [80], for instance autobiographical memories. The only study to report a high level of correlation between subjective and objective cognitive function examined the relationship between a simple question about global memory (GSE‐My) and an objective (or observer‐based) measure of autobiographical memory (CUAMI) [63]. However, this correlation may be much higher because a large change in cognitive function had just been induced by ECT. Complex scales aiming to measure individual domains of subjective cognition may ask questions in a way that is difficult to understand, contributing to a mismatch between subjective difficulties and objective performance.

Our meta‐analysis demonstrated that all but one of the subjective cognition measures found similar, weak correlations between subjective and objective cognition. Domain‐specific subjective measures (e.g., Subjective Memory Complaints Scale [SMCS]) did not demonstrate stronger correlations with objective cognitive function than global subjective measures. While these sub‐analyses were based on small sample sizes, this is in keeping with previous work examining the internal structure of the COBRA, which deemed a one‐factor structure the most appropriate [18]. Therefore, global assessment is a reasonable approach to measuring subjective cognition and may provide the same insight as a more intensive measure. Future studies could include a global question regarding cognitive function and compare this with more detailed scales.

4.2. Clinical Characteristics

Negative cognitions associated with depression may lead to patients perceiving that they have higher levels of cognitive problems. It has been argued that cognitive complaints represent depression severity, and thus, more negativity bias, rather than objective cognitive dysfunction [19, 76]. Our meta‐analysis showed that mood state did not have a significant impact on the correlation between subjective and objective cognition. However, 7 of 10 studies examining the discrepancy found that higher depressive symptoms [21, 49, 51, 62, 70] or more severely depressed participants [48, 54] were associated with underestimation of cognitive performance. This is because correlational analyses include patients with both over‐ and underestimation of their cognitive function whereas examination of the discrepancy allows for a more individualised, nuanced approach. Therefore, mood state may moderate this relationship; however, our meta‐analysis of correlational analyses was not able to demonstrate this. This warrants further investigation through meta‐analysis of discrepancy studies; however, this is currently premature due to the small number of studies.

4.3. Methodological Limitations of Current Evidence

Only 38% of studies adjusted for the impact of confounders on the association between subjective and objective cognition (e.g., IQ, age, medications and mood symptoms). This is particularly significant in the case of IQ, which is of course associated with objective cognitive performance [81]. Four studies examined the effect of IQ on the discrepancy between subjective and objective cognitive function, all finding a significant association. However, the direction of this relationship was not consistent and may depend on which domain of cognition is examined and the mood state of patients. For example, while lower IQ was associated with underestimation of global cognitive performance [70] and attention and processing speed [70], higher IQ was associated with underestimation of cognitive performance in attention and memory [49], and working memory and executive function in two studies [36, 62], but overestimation in another [70]. Given the conflicting findings and small number of studies examining this, it is difficult to draw a clear conclusion. It is possible that, because of greater cognitive reserve, individuals with higher IQ may experience more subjective impairment before objective deficits are able to be demonstrated on objective testing (i.e., there is more cognitive function to lose in order to cross into the ‘impaired range’ on a cognitive test). On the other hand, due to higher cognitive reserve, individuals with higher IQ may be less likely to develop objective cognitive impairment with a similar cognitive insult. They may have more sophisticated cognitive coping strategies to manage deficits both in daily life and in testing, leading to lower levels of perceived and measured deficits. A recent study of 111 adults with MDD found that while only 25% of participants were classified as cognitively impaired using normative comparisons, 62.2% were classified as idiographically impaired (i.e., compared with premorbid estimates) [25]. Thus, comparison to IQ and premorbid cognitive function can impact greatly on who is classified as ‘impaired’ [7], and is essential in studies examining cognitive functioning.

Further, clinical and demographic factors such as age, gender, and comorbidity may contribute to the relationship between subjective and objective cognitive function. Our meta‐analysis was not able to examine the impact of these factors in a meaningful way given the analyses were based on study rather than patient level data. Both psychiatric (e.g., anxiety, Attention Deficit Hyperactivity Disorder [ADHD]) and physical (e.g., hypertension, non‐insulin dependent diabetes mellitus) comorbidities are associated with cognitive impairment [82, 83]. Many studies did not collect data on these.

4.4. Strengths and Limitations

This review is topical and extremely relevant to the design of trials attempting to understand and treat cognitive impairment in mood disorders. Our meta‐analysis examined the association between subjective and objective cognition across a large number of participants (n = 4630). However, there are limitations that should be acknowledged when interpreting the results. There was significant heterogeneity between studies meaning that while meta‐analysis was possible, subgroup analyses were based on small numbers of studies. Patient samples were heterogenous with respect to clinical characteristics such as mood symptom severity and medication use. While most studies were of good quality, many did not adjust for confounders, including IQ. Just five studies were in unmedicated patients [38, 44, 63, 71, 74]. Medication may have impacted on cognition in the remaining 50 studies [84, 85] and was not able to be accounted for in the meta‐analysis. Similarly, medication for physical health problems may contribute to the association; however, it was not able to be accounted for in our meta‐analysis. A small number of studies did not provide mean scores for mood rating scales, meaning it was not possible to examine the impact of mood state on cognitive function in every case. We were unable to examine the effect of potential confounders such as age, gender, and comorbidities. The nature of the results available allowed us to undertake a meta‐analysis including studies examining the correlation between subjective and objective cognition. However, this does not take the direction of the discrepancy into account, including patients with both under‐ and overestimation of their cognition.

5. Conclusions

The results of this systematic review and meta‐analysis demonstrate a statistically significant but weak relationship between global measures of subjective and objective cognitive function in mood disorders, and between all cognitive domains except sustained attention. While this result supports the general consensus that there is not a strong relationship between subjective and objective cognitive function in mood disorders, this is based on correlational analyses which have the limitation of including patients with both under‐ and overestimation of their cognition. There were some novel and potentially impactful findings, including the lack of significant impact of remission status on the correlation between subjective and objective cognition. The choice of subjective measure did not impact significantly on the strength of correlation between subjective and objective cognitive function. Many of the studies did not allow for examination of the effects of confounders including IQ, mood symptoms, medications and age. This likely impacted on our findings and represents a significant gap in our current understanding of cognitive function in mood disorders.

Our finding of a weak association between objective and subjective cognitive function has important clinical implications. For example, a person with MDD and comorbid diabetes likely has a very different subjective experience of their cognition compared with a person with mania and a high level of grandiosity. Therefore, in clinical practice, we recommend assessing both objective and subjective cognitive function to adequately evaluate cognitive status and determine management. The finding of a significant discrepancy may lead the clinician to re‐evaluate the diagnosis and pursue further investigations. For instance, in the scenario of severe objective but minimal subjective impairment, the possibility of dementia would need to be evaluated. Alternatively, very significant subjective impairment with no evidence of objective impairment could potentially relate to a functional cognitive disorder [86].

Given the weak association between subjective and objective cognitive function, we suggest that future studies of cognition in mood disorders include measures of both subjective and objective cognitive function and assess the relationship between these outcomes. An assessment of the discrepancy, rather than correlation, between subjective and objective cognition allows for a more nuanced and useful interpretation of this relationship. A more refined measure of cognition may allow for a more useful assessment of cognition in both research and clinical practice. For instance, a combined measure including both objective and subjective cognition would assist in understanding this relationship in a more nuanced way. It would be useful to examine the effect of likely confounders on this association, particularly IQ, mood state, age, medications, comorbidities and substance abuse. A measure incorporating these factors may improve the ecological validity of cognitive testing, allowing better understanding of cognition both in research and clinical settings.

Conflicts of Interest

K.D. and R.P. report use of software for Cognitive Remediation interventions at no cost for research from SBT‐Pro. R.P. reports receiving support for travel to educational meetings from Servier and Lundbeck. K.W.M. reports having received consultancy fees from Lundbeck, Gedeon Richter, and Angelini in the past 3 years.

Supporting information

Figure S1: Effect of diagnosis on relationship between subjective and objective cognition.

Figure S2: Effect of domain‐specific subjective cognitive measures on relationship between subjective and objective cognition.

Figure S3: Effect of specific subjective measures on relationship between subjective and objective cognition.

Table S1: Neuropsychological domains for meta‐analysis.

Table S2: Examination of factors impacting relationship between subjective and objective cognitive function (n = 20).

Table S3: Meta‐analytic findings for relationship between specific subjective measures and global objective cognition.

Eggleston K., Miskowiak K. W., Porter R., Frampton C., and Douglas K., “Systematic Review and Meta‐Analysis of the Association Between Subjective and Objective Cognitive Function in Mood Disorders,” Bipolar Disorders 28, no. 1 (2026): e70077, 10.1111/bdi.70077.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.