Acta Psychiatr Scand 2013: 1–14 © 2013 John Wiley & Sons A/S.

2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
All rights reserved ACTA PSYCHIATRICA SCANDINAVICA
DOI: 10.1111/acps.12133

Meta-analysis

Neuropsychological testing of cognitive

impairment in euthymic bipolar disorder: an
individual patient data meta-analysis
Bourne C, Aydemir O, € Balanz a-Martı́nez V, Bora E, Brissos S, € Aydemir3, V.
C. Bourne1,2,*, O.
Cavanagh JTO, Clark L, Cubukcuoglu Z, Videira Dias V, Dittmann S, Balanzá-Martínez4, E. Bora5,
Ferrier IN, Fleck DE, Frangou S, Gallagher P, Jones L, Kiesepp€a T, S. Brissos6,7, J. T. O. Cavanagh8,
Martı́nez-Aran A, Melle I, Moore PB, Mur M, Pfennig A, Raust A, L. Clark9, Z. Cubukcuoglu10,
Senturk V, Simonsen C, Smith DJ, Bio DS, Soeiro-de-Souza MG, V. V. Dias11, S. Dittmann12,
Stoddart SDR, Sundet K, Sz€ oke A, Thompson JM, Torrent C, Zalla T,
Craddock N, Andreassen OA, Leboyer M, Vieta E, Bauer M,
I. N. Ferrier13, D. E. Fleck14,
Worhunsky PD, Tzagarakis C, Rogers RD, Geddes JR, Goodwin GM. S. Frangou15, P. Gallagher13,
Neuropsychological testing of cognitive impairment in euthymic bipolar L. Jones16, T. Kiesepp€a17,18,
disorder: an individual patient data meta-analysis. A. Martínez-Aran19,*,
I. Melle20,21, P. B. Moore13,
Objective: An association between bipolar disorder and cognitive M. Mur22, A. Pfennig23,*,
impairment has repeatedly been described, even for euthymic patients. A. Raust24,*, V. Senturk25,
Findings are inconsistent both across primary studies and previous C. Simonsen20,26, D. J. Smith27,
meta-analyses. This study reanalysed 31 primary data sets as a single D. S. Bio28, M. G. Soeiro-de-
large sample (N = 2876) to provide a more definitive view. Souza28, S. D. R. Stoddart16,
Method: Individual patient and control data were obtained from
original authors for 11 measures from four common
K. Sundet20,26, A. Sz€oke24,29,
neuropsychological tests: California or Rey Verbal Learning Task J. M. Thompson13, C. Torrent19,*,
(VLT), Trail Making Test (TMT), Digit Span and/or Wisconsin Card T. Zalla30, N. Craddock27,
Sorting Task. O. A. Andreassen20,21,*,
Results: Impairments were found for all 11 test-measures in the bipolar M. Leboyer24,29,*, E. Vieta19,*,
group after controlling for age, IQ and gender (Ps  0.001, M. Bauer23,*, P. D. Worhunsky1,
E.S. = 0.26–0.63). Residual mood symptoms confound this result but C. Tzagarakis1, R. D. Rogers1, J.
cannot account for the effect sizes found. Impairments also seem R. Geddes1,*, G. M. Goodwin1,*
unrelated to drug treatment. Some test-measures were weakly
correlated with illness severity measures suggesting that some Key words: bipolar disorder; cognitive impairment;
impairments may track illness progression. review; neuropsychological tests
Conclusion: This reanalysis supports VLT, Digit Span and TMT as Corin Bourne, Department of Psychiatry, University of
robust measures of cognitive impairments in bipolar disorder patients. Oxford, Warneford Hospital, Oxford OX3 7JX, UK. E-
The heterogeneity of some test results explains previous differences in mail: [email protected]
meta-analyses. Better controlling for confounds suggests deficits may be
*European Network of Bipolar Research Expert Centres.
smaller than previously reported but should be tracked longitudinally
across illness progression and treatment.

Accepted for publication March 1, 2013

Summations
• Cognitive deficits are present in euthymic bipolar patients, and although some confounds may
explain part of the previously reported effect sizes, they cannot entirely explain the impairments.
• Individual patient data meta-analysis has important advantages over the use of published summary
data for systematic review especially with regard to controlling for confounds.

1
Bourne et al.

Considerations
• The relative lack of drug effects on neuropsychological test performance should be treated with
caution as this mega-analysis could not take into account duration or dosage of each drug treatment.
• Similarly, the correlational analysis suggesting that some impairments may track illness progression
should also be treated with caution until longitudinal data supports the causality of this relationship.

1
Department of Psychiatry, University of Oxford, Oxford, 2Department of Psychology & Counselling, Newman University, Birmingham, UK, 3Department of Psychiatry, Celal Bayar
University, Manisa, Turkey, 4Service of Psychiatry, University Hospital Doctor Peset, University of Valencia, CIBERSAM, Valencia, Spain, 5Department of Psychiatry, University of
Melbourne, Melbourne, Vic., Australia, 6Lisbon’s Psychiatric Hospitalar Centre, Lisbon, Portugal, 7Janssen Pharmaceutical, Queluz de Baixo, Portugal, 8Department of
Psychological Medicine, University of Glasgow, Glasgow, 9Department of Psychology, University of Cambridge, Cambridge, UK, 10Department of Child and Adolescent Psychiatry,
LVR-Kliniken Bedburg-Hau, Bedburg-Hau, Germany, 11Bipolar Disorder Research Program, Faculty of Medicine, Hospital Santa Maria, University of Lisbon (FMUL), Lisbon,
Portugal, 12Department of Psychiatry, Ludwig-Maximilians-Universität, Munich, Germany, 13Institute of Neuroscience, Newcastle University, Newcastle upon Tyne, UK,
14
Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati Academic Health Center, Cincinnati, OH, USA, 15Mount Sinai School of Medicine, NY, USA,
16
Department of Psychiatry, School of Clinical & Experimental Medicine, University of Birmingham, Birmingham, UK, 17Department of Mental Health and Substance Abuse
Services, National Institute for Health and Welfare, Helsinki, 18Department of Psychiatry, Peijas Hospital, Helsinki University Central Hospital, Vantaa, Finland, 19Institute of
Neurosciences, Hospital Clinic, University of Barcelona, IDIBAPS, CIBERSAM, Barcelona, Catalonia, Spain, 20Division of Mental Health and Addiction, Oslo University Hospital,
Ulleval, Oslo, 21Institute of Clinical Medicine, University of Oslo, Oslo, Norway, 22Psychiatry Service, Santa Maria Hospital, University of Lleida, IRBLleida, Lleida, Spain,
23
Department of Psychiatry and Psychotherapy, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany, 24AP-HP, Service de psychiatrie,
Hôpital H. Mondor – A. Chenevier, Creteil, France, 25Department of Psychiatry, Ankara University School of Medicine, Ankara, Turkey, 26Department of Psychology, University of
Oslo, Oslo, Norway, 27Department of Psychological Medicine and Neurology, Cardiff University School of Medicine, Cardiff, UK, 28Mood Disorders Unit (GRUDA), School of
Medicine, University of Sao Paulo (HC-FMUSP), Sao Paulo, Brazil, 29Inserm, U955, Equipe 15, Creteil and 30Institut Jean Nicod? CNRS, Ecole Normale Supérieure, Paris, France

patients: Arts et al. (6); Bora et al. (7); Robinson

Introduction
et al. (8) and Torres et al. (9). A fifth paper, Rob-
Bipolar disorder has been associated with cognitive inson and Ferrier (10), provided a narrative review
impairment even in euthymia (1–4). Specific of studies that considered the relationship between
domains of impairments include the following: illness variables and cognitive deficits. Surpris-
executive control (verbal and category fluency, ingly, despite the similar aims, similar search terms
mental manipulation, set shifting, response inhibi- and overlapping databases used across the five
tion), verbal learning and memory, visual memory papers, they demonstrate wide variation in the pri-
and attention (5–12). A subset of such deficits may mary studies chosen for inclusion and in their spe-
also be present in first-degree relatives of bipolar cific conclusions. An additional meta-analysis was
patients indicating a possible endophenotype for published in 2011: Mann-Wrobel et al. (12). This
bipolar disorder (6, 7) and a starting point for paper differed from three of the earlier meta-analy-
further genetic understanding of the disorder. ses (6, 8, 9) by not supporting a differential impair-
Some studies (10) have reported a correlation ment in verbal memory and executive function.
between a subset of cognitive decrements and ill- Kurtz & Gerraty (13) provided a further meta-
ness history, suggesting the competing possibility analysis although they considered clinical groups
that clinical episodes may cause impairments in the other than purely euthymic patients. If this study is
neuronal function relevant to these domains. Such restricted to euthymic patients, then their meta-
acquired effects might be more amenable to analysis had similar sample sizes and effect size
improved treatment. Given the important potential range to the meta-analyses reviewed in more detail
implications for the neurobiology of bipolar disor- here (see Table 1). However, the authors also sug-
der and its treatment, it is disappointing that these gested a greater impairment was present for verbal
claims rest on studies of modest size that allow lit- memory but also for non-verbal delayed memory
tle confidence in their individual conclusions. Rep- in contrast to other meta-analyses (6, 7).
lication is confused by the adoption of a too wide The present study provides an independent indi-
range of different neuropsychological tests with vidual patient data meta-analysis (IPDMA) of the
varying sensitivity to and specificity for particular data from the most comparable studies included in
cognitive domain impairments and their neural the previous reviews, given the authors could pro-
substrates. vide original data for pooling. IPDMA has not
Between 2006 and 2010, four papers have con- been widely used in psychiatry although it is
ducted meta-analyses on the cognitive deficits asso- increasingly used in medical genetics (14–16) where
ciated with bipolar disorder in purely euthymic it is often termed ‘mega-analysis’. It has important

2
 Euthymic bipolar cognition: IPDMA

advantages over the use of published summary cessing speed; iii) Digit Span [from WAIS-R Digit
data for systematic review (17). In particular, IP- Span (21)] as a non-word working memory span
DMA allows the primary study effect sizes to be task and iv) Wisconsin Card Sorting Task (WCST)
adjusted for confounding factors (i.e. factors such (22) as a measure of set shifting and rule discovery.
as age, education and IQ) prior to meta-analysis Verbal Learning Task (VLT), TMT and WCST all
and for a large data set to be analysed for drug and appear in the International Society for Bipolar
illness severity effects. The latter having been previ- Disorders recently recommended battery for neu-
ously restricted to primary studies of modest ropsychological assessment (23).
sample size or narrative review. The adjustment From the four selected neuropsychological tests,
for confounding factors is especially valuable we focused on 11 specific outcome measures: VLT
because, although some of the primary studies total score on trials 1–5 (Total1–5), VLT score on
were very tightly matched case–control studies Short Delay (ShortDelay), VLT score on Long
focusing on one or two neuropsychological tests, Delay (LongDelay), VLT score on Recognition
other included studies were more opportunistic (Recognition), VLT score for Recognition minus
samples running large neuropsychological test bat- score for False Positives (Recog-FP); time to com-
teries with more sample variation. In a standard plete Trail Making Test A (TMTA), time to com-
meta-analysis, the results from these two types of plete Trail Making Test B (TMTB); score on
study are combined without adjustment. Forward Digit Span (FDS), score on Reverse Digit
Span (RDS); number of categories achieved on
Wisconsin Card Sorting Task (WCSTCats.) and
Aims of the study
number of perseverations on Wisconsin Card
The main aim of the study was to synthesize data Sorting Task (WCSTPersev.).
demonstrating cognitive deficits in euthymic bipo- Where possible, demographic and clinical vari-
lar patients in such a way as to be able to adjust ables were also collected for each primary data
for confounding factors to provide a more defini- set including i) age; ii) IQ; iii) current mood; iv)
tive estimate for effects sizes than in prior meta- age at onset; v) number of prior manic and
analyses. A secondary aim was to create a large depressed episodes; vi) number of prior manic
data set to provide a more definitive view of drug and depressed hospitalizations and vii) drug
and illness severity effects on cognitive impair- treatment history.
ments than has been possible in relative small sam-
ple primary studies. We chose to include tests that
Search strategy
had appeared consistently in the meta-analyses as
showing impairment and for which data were actu- Given the existence of five recent prior reviews in
ally available for the majority of individual this area (each with similar but different search
patients. terms and inclusion/exclusion criteria), this study
did not conduct an additional full systematic
search under PRISMA (24) rules. Rather, in an
Material and methods
attempt to include all the primary studies that had
Table 1 shows the results from the four existing been in the prior reviews, all first authors of studies
meta-analyses as the rank of the neuropsychologi- appearing in the five review papers that contained
cal tests showing the largest effects in each review. data on at least one of the four required neuropsy-
Effect sizes appear to be relatively large, but it is chological tests were contacted. In addition, Psy-
striking that sample numbers vary considerably chInfo and PubMed databases were searched with
due to the differences in criteria for study inclu- the key concepts of bipolar disorder, euthymia and
sion. The relative order of neuropsychological tests cognitive impairment to find any additional pri-
when ranked by effect size is variable from analysis mary studies that met our criteria. These searches
to analysis partly due to the variation in study were restricted to articles published between 1 Jan-
inclusion and probably partly due to noise. uary 2007 and 30 June 2010 in English language
Primary data were sought that tested both eu- peer-reviewed journals. In total, 45 primary studies
thymic bipolar patients and healthy controls (aged were identified from 41 different authors (see
18–65) on at least one of four key neuropsycholog- Table S1). This number is smaller than may have
ical tasks identified in Table 1: i) a verbal learning first appeared from the literature search as some
and memory task, that is, California Verbal Learn- studies incorporated data sets used in other pub-
ing Task (CVLT) (18) or Rey Verbal Learning lished studies and therefore did not constitute
Task (RAVLT) (19); ii) the Trail Making Test mutually exclusive data sets. Of the 45 eligible pub-
(TMT) (20) as a measure of set shifting and pro- lished studies, full data were provided by primary

3
Bourne et al.

Table 1. Summary of the effect sizes found for neuropsychological performance of bipolar patients relative to healthy controls. Top seven effect sizes in the meta-analysis by (a)
Arts et al. (6), (b) Bora et al. (7), (c) Robinson et al. (8) and (d) Torres et al. (9)

Neuropsychological test Cognitive domain N (bipolar) N (control) Effect size P

(a)
1 RDS Executive 222 205 1.02 <0.0001
2 TMTB Executive 309 306 0.99 <0.0001
3 WCST (Perseveration) Resp. Inhib 268 288 0.88 <0.0001
4 Category Fluency Executive 178 178 0.87 <0.0001
5 Rey/CVLT (Delayed Recall) Verb. L + M 269 282 0.85 <0.0001
6 Digit Symbol Subtest Attention 202 249 0.84 <0.0001
7 Rey/CVLT (Total Recall) Verb. L + M 369 382 0.82 <0.0001
(b)
1 TMTB Executive 793 626 0.86 <0.0001
2 Rey/CVLT (Learning) Verb. L + M 619 632 0.85 <0.0001
3 CPT Omission Attention 303 279 0.83 <0.0001
4 Rey/CVLT (Delayed Recall) Verb. L + M 578 612 0.77 <0.0001
5 Stroop Resp. Inhib 746 707 0.76 <0.0001
6 Digit Symbol Subtest Attention 381 479 0.75 <0.0001
7 RDS Executive 375 487 0.75 <0.0001
(c)
1 Category Fluency Executive 149 135 1.09 <0.0001
2 RDS Executive 222 209 0.98 0.0031
3 Rey/CVLT (Total Recall) Verb. L + M 344 347 0.90 <0.0001
4 TMTB Executive 418 355 0.78 <0.0001
5 WCST (Perseveration) Resp. Inhib 195 216 0.76 <0.0001
6 Rey/CVLT (Short Free Recall) Verb. L + M 345 349 0.73 <0.0001
7 Rey/CVLT (Long Free Recall) Verb. L + M 365 368 0.71 <0.0001
(d)
1 Rey/CVLT (Total Recall) Verb. L + M 381 439 0.81 <0.0001
2 Digit Symbol Subtest Attention 222 310 0.79 <0.0001
3 Rey/CVLT (Short Delay) Verb. L + M 315 307 0.74 <0.0001
4 CPT Hits Attention 188 208 0.74 <0.0001
5 Rey/CVLT (Long Delay) Verb. L + M 361 441 0.72 <0.0001
6 Stroop Resp. Inhib 346 329 0.71 <0.0001
7 WCST (Perseveration) Resp. Inhib 244 229 0.69 <0.0001

CPT, continuous performance task; CVLT, California Verbal Learning Task; RDS, Reverse Digit Span; Resp. Inhib, Response Inhibition; TMTB, Trail Making Test B; Verb. L + M,
Verbal Learning and Memory; WCST, Wisconsin Card Sorting Task.

authors in relation to 25 published papers (4, cian Administered Rating Scale for Mania Factor
25–48) with the data from the remaining 20 eligible 1 CARS-M(F1) (55) or  20 on Manic State Rat-
studies unavailable and therefore not included in ing Scale (MSRS) (56). If no mood ratings were
this reanalysis. Additionally, new primary data available, then euthymia had been assessed by a
that met our criteria were also provided in relation qualified psychiatrist only.
to six unpublished data sets (49, 50) (A. Macritche, The total sample size for the reanalysis was
manuscript in preparation; A. Varma, manuscript therefore 2876 participants: 1276 euthymic bipolar
in preparation; A. Pfennig, M. Alda, T. Young, patients (54.7% female) and 1609 healthy controls
G. MacQueen, J. Rybakowski, A. Suwalska, (53.5% female). The bipolar patients were 83.5%
C. Simhandl, B. König, T. Hajek, C. O‘Donovan, Bipolar I, 12.3% Bipolar II, 2.7% Bipolar NOS,
S. von Quillfeldt, D. Wittekind, J. Ploch, C. Sauer, 1.4% Schizoaffective Disorder.
M. Bauer, manuscript in preparation; M.G. Soeiro-
de-Souza & D. Soares-Bio, manuscript in
Statistical analyses
preparation), giving a total of 31 primary data sets
for this reanalysis as shown in Table 2. Parametric statistical tests were used to compare a
Where mood scores were available, euthymia variety of demographic variables between bipolar
was defined as  8 on Hamilton Depression Rat- patients and healthy controls. Where appropriate,
ing Scale (HDRS) (51) or  15 on Montgomery– homogeneity of variance was checked using
Asberg Depression Rating Scale (MADRS) (52) or Levene’s test. All continuous measures (including
 11 on Inventory of Depressive Symptomatology depression and mania scores) were converted to
(Clinician Rating; IDS-C) (53) and  8 on Young standardized z-scores within each study sample
Mania Rating Scale (YMRS) (54) or  8 on Clini- (patients plus controls) before further analysis.

4
 Euthymic bipolar cognition: IPDMA

Group effect size of cognitive deficits. To investigate Table 2. List of studies in reanalysis data set
group (patient vs. control) effects on neuropsycho- Study N Nbp Ncont
logical performance, group, age, IQ and gender
were regressed on to each of the 11 neuropsycho- 1 Balanza-Martinez 41 15 26
et al. (26)
logical test outcome measures within each of the
2 Bora et al. (27) 95 65 30
31 studies. For the eight studies that did not use an 3 Cavanagh et al. (28) 39 19 20
explicit measure of IQ, years of education was used 4 Clark et al. (29) 60 30 30
as a proxy (rp = 0.50, P < 0.001). The regression 5 Cubukcuoglu & 101 51 50
Aydemir (49)
coefficient and standard error for group within 6 Dias et al. (46) 115 65 50
each study were then entered for meta-analysis for 7 Dittmann et al. (30) 116 74 42
each outcome variable. Thus, the meta-analysis 8 El-Badri et al. (31) 57 30 27
was effectively performed on study group effect 9 Fleck et al. (32) 51 11 40
10 Fleck et al. (33) 70 22 48
sizes adjusted a priori for the confounds of age, IQ 11 Frangou et al. (34) 86 42 44
and gender. The meta-analyses were conducted on 12 Goswami et al. (35) 74 37 37
both fixed and random effects assumptions, but 13 Hellvin et al. (50)* 228 63 165
results did not differ materially. This analysis did 14 Kaya et al. (48) 62 43 19
15 Kieseppa et al. (36) 140 26 114
not use the more standard IPDMA technique of 16 A. Macritche 56 28 28
mixed model regression (with fixed and random (manuscript in
effects) as the between-study heterogeneity for preparation)
group effect size was considered too high for at 17 Martinez-Aran et al. (4) 69 39 30
18 Martinez-Aran et al. (37) 112 77 35
least some of the outcome measures (see Table 4). 19 Mur et al. (38) 89 43 46
20 A. Pfennig, M. Alda, 54 33 21
Residual mood effects. Residual mood symptoms T. Young, et al.
(manuscript in preparation)
(both depression and mania) could not be added to
21 Senturk et al. (39) 56 27 29
the above analysis because they were confounded 22 Simonsen et al. (25) 146 29 117
with group. However, in an attempt to understand 23 Simonsen et al. (47)† 204 31 173
how much of the group effect on performance 24 Smith et al. (40) 54 21 33
25 M.G. Soeiro-de-Souza & 134 38 96
might be attributable to residual confounding by D. Soares-Bio
mood, two further analyses were conducted. The (manuscript in preparation)
first approach used meta-regression, with each of 26 Stoddart et al. (41) 59 19 40
the studies ascribed a factor relating to the relative 27 Szoke et al. (42) 145 97 48
28 Thompson et al. (43) 126 63 63
level of residual mood symptoms in the patient 29 Torrent et al. (44) 73 38 35
group. The second method considered mood 30 A. Varma 106 53 53
effects within the patient group only using mixed (manuscript in preparation)
model regression with data collapsed across stud- 31 Zalla et al. (45) 58 38 20

ies. Depression scores and mania scores along with Grand total 2876 1267 1609
age, IQ, gender (all fixed effects) and study (ran-
*Data set reduced from that published to exclude participants already included in
dom effect) were regressed on to each of the 11 Simonsen et al. (25, 47).
neuropsychological test outcome measures. †Data set reduced from that published to exclude participants already included in
Simonsen et al. (25).
Drug effects within patient group. To investigate
potential drug effects within the patient group, ber of depressed episodes, number of manic epi-
mixed model linear regression was used. Patients sodes, total number of episodes, number of
were coded for five binary (yes/no) drug status depressed hospitalizations, number of manic hos-
variables: lithium, anticonvulsants, antipsychotics, pitalizations, total number of hospitalizations and
antidepressants and drug free. Each drug status illness duration were each fitted separately into the
variable (fixed effect) together with age, IQ, gender regression model with age, IQ and gender as uni-
(fixed effects) and study (random effect) was versal confounders (fixed effect) and study (ran-
regressed on to each of the 11 neuropsychological dom effect) for each of the 11 neuropsychological
test outcome measures. test outcome measures.
Statistical analysis was conducted in R 2.12.2
Relationship between illness variables and cognitive (The R Foundation for Statistical Computing,
deficits. Mixed model linear regression was also Vienna, Austria) except for the meta-analysis
used to investigate potential relationships between which was conducted in STATA IC Version 11
illness severity measures and neuropsychological (StataCorp LP., College Station, TX, USA). All
test performance within the patient group. Num- statistical tests were two-tailed.

5
Bourne et al.

Table 3. Demographics of patient and control groups

Results
Patients (N = 1267) Controls (N = 1609)
Table 3 shows the demographic profile of the N = 2876 M (SD) M (SD)
patient and control groups. Overall, the groups
Age (n = 2868)
were well matched for gender (v21 = 0.71, P = 0.40) Male
38.8 (11.7)
569
36.4 (11.8)
748
but showed a significant difference in age Female 698 861
(t2866 = 5.51, P < 0.001, d = 0.21; 95% CI, Years of education (n = 2716) 12.9 (3.4) 13.6 (3.0)
1.57–3.30) with bipolar patients being, on average, IQ measures
NART/WAIS (n = 1103) 112.4 (11.6) 114.7 (10.6)
2.4 years older. The bipolar group also had, on WASI (n = 961) 107.5 (10.5) 112.6 (10.8)
average, 0.6 fewer years of education (t2714 = 5.14, WMS-R (n = 100) 97.4 (17.8) 105.0 (14.8)
P < 0.001, d = 0.20; 95% CI, 0.88 to 0.40) and WAIS Vocab. Subtest (n = 181) 44.6 (11.5) 47.1 (12.0)
showed a difference in premorbid IQ on the two WAIS Info. Subtest (n = 56) 19.4 (5.0) 21.4 (3.9)
IQ measures with substantial sample sizes: NART, National Adult Reading Test; WAIS, Wechsler Adult Intelligence Scale;
National Adult Reading Test (NART) (57)/WAIS- WASI, Wechsler Abbreviated Scale of Intelligence; WMS-R, Wechsler Memory
R (21) (t985 = 3.87, P < 0.001, d = 0.25; 95% CI, Scale.
3.86 to 1.26) and Wechsler Abbreviated Scale
of Intelligence (WASI) (58) (t959 = 6.99, P < 0.001,
d = 0.48; 95% CI, 6.61 to 3.71). The groups and 15% for VLT Recog-FP can be considered
did not differ on IQ for those studies that used the minor; values of 39% (VLT ShortDelay), 42%
WAIS Vocabulary Subtest (18) (t179 = 1.2, (VLTLlongDelay FDS) and 45% (WCSTPersev.)
P = 0.23; 95% CI, 6.65 to 1.66) or WAIS Infor- can be considered moderate; whilst between-study
mation Subtest (21): t54 = 1.7, P = 0.10; 95% CI, heterogeneity on VLT Total1–5 (Fig. 1), TMTB
4.41 to 0.41). One study used the Wechsler Mem- (Fig. 2) and both FDS and RDS with I2 = 61%,
ory Scale (WMS-R) (59) as an IQ measure which 69%, 71% and 84%, respectively, was substantial
showed a group difference (t98 = 2.31, P = 0.02, (60). Magnitude of effect size was associated with
d = 0.46; 95% CI, 14.1 to 1.1) but as this is a increased heterogeneity.
memory measure and not a measure of premorbid
IQ, this difference is not surprising. It should be
Residual mood effects
noted that the last three measures were only used
in relatively small sample subsets. Overall, the data The meta-regression showed that the factor relat-
set showed significant group differences in a range ing to a study’s ability to minimize residual mood
of confounding variables reinforcing the need to within the patient group significantly explained the
covary for these factors in any combined analysis. between-study heterogeneity for two of the 11 out-
This can only be done convincingly using IPDMA. come variables: TMTA regression coefficient =
0.05 (t = 2.18, P = 0.047, Adj.R2 = 100%, 95%
CI, 0.10 to 0.001) and WCSTCats. regression
Group effect size of cognitive deficits
coefficient = 0.07 (t = 2.78, P = 0.018,
The patient group had large reductions in perfor- Adj.R2 = 100%, 95% CI, 0.12 to 0.02). None
mance on all 11 outcome variables relative to con- of the other nine outcome variables were associ-
trols when controlling for the effect of age, IQ and ated with significant meta-regression coefficients
gender. The overall effect size for group varied (VLT Total1–5: t = 0.70, P = 0.50, 95% CI, 0.08
between 0.63 on TMTB to 0.26 on WCSTCats. to 0.04; VLT ShortDelay: t = 1.10, P = 0.29, 95%
(Table 4). The sample sizes (n in Table 4) were CI, 0.07 to 0.02; VLT LongDelay: t = 0.82,
substantially larger than for the meta-analyses in P = 0.42, 95% CI, 0.07 to 0.03; VLT Recogni-
Table 1a,c,d and comparable to or larger than tion: t = 0.18, P = 0.86, 95% CI, 0.04 to 0.05;
Table 1b. Figures 1–3 show forest plots for the VLT Recog-FP: t = 1.43, P = 0.19, 95% CI, 0.08
meta-analysis of the confound-adjusted group to 0.02; TMTB: t = 0.72, P = 0.48, 95% CI, 0.12
effect sizes associated with VLT, TMT and WCST to 0.06; FDS: t = 0.07, P = 0.94, 95% CI, 0.09 to
neuropsychological tests. 0.08; RDS: t = 0.20, P = 0.84, 95% CI, 0.89 to
The studies showed a wide range of between- 0.11; WCSTPersev.: t = 1.56, P = 0.15, 95% CI,
study heterogeneity across the 11 outcome mea- 0.02 to 0.13).
sures, ranging from 0% to 84% (Table 4). The I2 The second approach to understand residual
measure of heterogeneity provides an indication of mood effects was to consider the effect of depres-
the proportion of total variation in effect size esti- sion score and mania score on neuropsychological
mates attributable to between-study heterogeneity. performance within the patient group only.
I2 values of 8% for TMTA, 12% for WCSTCats. Depression score showed an overall main effect on

6
 Euthymic bipolar cognition: IPDMA

Table 4. Overall effect size of group for the 11 outcome variables

Outcome Overall effect

Test variable N (bipolar) N (control) size (95% CI) P I2 (%)

VLT Total1–5 624 661 0.51 (0.42–0.60) <0.001 61

VLT Short Delay 667 680 0.48 (0.39–0.57) <0.001 39
VLT Long Delay 667 680 0.55 (0.47–0.64) <0.001 42
VLT Recognition 576 590 0.46 (0.36–0.57) <0.001 0
VLT Recog-FP 333 404 0.38 (0.26–0.50) <0.001 15
TMT A 879 752 0.49 ( 0.58 to 0.40) <0.001 8
TMT B 903 778 0.63 ( 0.72 to 0.55) <0.001 69
Digit Span Forward 533 650 0.30 (0.20–0.40) <0.001 71
Digit Span Reverse 533 650 0.60 (0.51–0.69) <0.001 84
WCST Categories 605 639 0.26 (0.15–0.37) <0.001 12
WCST Perseverations 606 639 0.29 ( 0.40 to 0.17) <0.001 45

Recog-FP, recognition minus false positives; TMT, Trail Making Test; VLT, Verbal Learning Task; WCST, Wisconsin Card Sorting Task.

just three of 11 outcome measures (when account- size = 0.29, P = 0.006, 95% CI, 0.49 to 0.08)
ing for the effect of mania, age, IQ and gender), of the 11 outcome measures (Ps > 0.1 for all other
typically on measures of memory, speed and execu- effect sizes of antipsychotic status except for VLT
tive function: VLT Total1–5 effect size = 0.09, ShortDelay and VLT LongDelay both with
t652 = 2.68, P = 0.008, 95% CI, 0.16 to 0.03; P = 0.08 and WCSTPersev. with P = 0.09). Being
VLT Recognition effect size = 0.13, t605 = 3.32, drug free improved performance (given effects of
P = 0.001, 95% CI, 0.02 to 0.05; and TMTA study, age, IQ and gender) relative to any drug on
effect size = 0.09, t682 = 2.62, P = 0.009, 95% CI, two of the 11 outcome measures: VLT Total1–5
0.02–0.16. Higher depression scores were related to (effect size = 0.39, P = 0.010, 95% CI, 0.69 to
worse cognitive performance but the effect size was 0.09) and VLT LongDelay (effect size = 0.35,
considerably smaller than the relevant effect size P = 0.017, 95% CI, 0.64 to 0.06; Ps > 0.1 for
for group (see Table 4). There was no overall main all other effect sizes of drug-free status).
effect of mania score on any of the 11 outcome
measures (when accounting for the effect of depres-
Relationship between illness variables and cognitive deficits
sion, age, IQ and gender).
Table 5 shows the illness characteristics of the
patient sample. The mixed model regression analy-
Drug effects within patient group
sis within the patient group suggested that some of
Within the patient sample, there was full informa- these illness variables correlated at better than
tion on drug treatment for 952 patients (75%) and chance with some of the 11 outcome variables
information on lithium status for 1122 (89%). (eight out of 66) but effects were generally small.
Thus, for comparative analysis, 652 patients were Thus, number of manic episodes affected perfor-
on lithium with 470 lithium free, 337 were on anti- mance on three of the outcome measures (given
convulsants with 409 anticonvulsant free, 209 were effects of study, age, IQ and gender): VLT Short-
on antidepressants with 537 antidepressant free, Delay (effect size = 0.07, P = 0.03, 95% CI,
209 were on antipsychotics with 537 antipsychotic 0.14 to 0.01); VLT LongDelay (effect
free and 72 were drug free compared to 880 on at size = 0.09, P = 0.007, 95% CI, 0.16 to 0.03);
least one drug type. The mixed model regression and TMTA (effect size = 0.09, P = 0.03, 95% CI,
analysis within the patient group suggested that 0.01–0.17). Number of total episodes only affected
neither lithium (given effects of study, age, IQ and performance on TMTA (effect size = 0.08,
gender) nor antidepressants (given effects of study, P = 0.03, 95% CI, 0.01–0.15). Number of depres-
age, IQ and gender) affected performance on any sive episodes had no main effects. Number of
of the 11 outcome measures (Ps > 0.1 for all effect depressive hospitalizations also only affected per-
sizes of lithium or antidepressant status). Similarly, formance on TMTA (effect size = 0.26, P = 0.003,
anticonvulsants showed no effect on performance 95% CI, 0.09–0.42) whilst number of total hospi-
(given effects of study, age, IQ and gender) on any talizations affected performance on TMTA (effect
of the 11 outcome measures (Ps > 0.1 for all effect size = 0.12, P = 0.008, 95% CI, 0.03–0.21), TMTB
sizes of anticonvulsants except for WCST Cats. (effect size = 0.13, P = 0.005, 95% CI, 0.04–0.21)
with P = 0.08). Antipsychotics (given effects of and WCSTCats. (effect size = 0.12, P = 0.01,
study, age, IQ and gender) showed a reduced per- 95% CI, 0.21 to 0.03). Number of manic hospi-
formance on VLT Total1–5 only (effect talizations had no main effects. Thus, of the four

7
Bourne et al.

(a) VLT Total1-5 (b) VLT ShortDelay

Study % Study %
ID ES (95% CI) Weight ID ES (95% CI) Weight

Bora et al 2007 0.48 (0.14, 0.82) 6.83 Bora et al 2007 0.62 (0.27, 0.97) 6.29
Cavanagh et al 2002 0.90 (0.36, 1.44) 2.69 Cavanagh et al 2002 0.80 (0.22, 1.39) 2.31
Clark et al 2002 0.75 (0.29, 1.21) 3.74 Clark et al 2002 0.58 (0.13, 1.02) 3.92
Cubukcuoglu & Aydemir 0.62 (0.27, 0.96) 6.53 Cubukcuoglu & Aydemir 0.54 (0.19, 0.88) 6.66
Fleck et al 2003 0.98 (0.40, 1.57) 2.30 Fleck et al 2003 0.76 (0.15, 1.37) 2.12
Goswami et al 2006 0.39 (–0.04, 0.82) 4.21
Goswami et al 2006 0.76 (0.36, 1.16) 4.87
Hellvin et al 0.22 (–0.07, 0.51) 9.43
Hellvin et al 0.05 (–0.23, 0.34) 9.86
Kaya et al 2007 0.98 (0.49, 1.46) 3.34
Kieseppa et al 2005 0.43 (0.06, 0.81) 5.55
Kieseppa et al 2005 0.44 (0.04, 0.84) 4.85
Martinez-Aran et al 2004 0.58 (0.16, 0.99) 4.64
Martinez-Aran et al 2004 0.47 (0.07, 0.88) 4.76
Martinez-Aran et al 2007 0.56 (0.20, 0.93) 5.85
Martinez-Aran et al 2007 0.34 (–0.02, 0.69) 6.22
Mur et al 2007 0.06 (–0.34, 0.46) 4.87
Mur et al 2007 –0.05 (–0.46, 0.35) 4.76
Pfenning et al –0.14 (–0.64, 0.36) 3.15
Pfenning et al 0.18 (–0.37, 0.72) 2.64
Simonsen et al 2008 0.02 (–0.35, 0.39) 5.73
Simonsen et al 2008 0.11 (–0.28, 0.50) 5.15
Simonsen et al 2011 0.55 (0.22, 0.89) 6.91
Simonsen et al 2011 0.39 (0.05, 0.72) 7.05
Smith et al 2006 1.00 (0.50, 1.50) 3.15 Smith et al 2006 0.88 (0.36, 1.39) 2.95
Stoddart et al 2007 0.80 (0.24, 1.35) 2.52 Stoddart et al 2007 0.86 (0.33, 1.40) 2.78
Thompson et al 2005 0.58 (0.28, 0.87) 8.85 Thompson et al 2005 0.53 (0.22, 0.84) 8.06
Torrent et al 2006 0.75 (0.36, 1.14) 5.06 Torrent et al 2006 0.52 (0.15, 0.88) 5.77
Varma et al 0.85 (0.51, 1.19) 6.91 Varma et al 0.84 (0.50, 1.18) 6.73
Overall (I-squared = 61.1%, p = 0.000) 0.51 (0.42, 0.60) 100.00 Overall (I-squared = 39.1%, p = 0.038) 0.48 (0.39, 0.57) 100.00

–1.57 0 1.57 –1.46 0 1.46

(c) VLT LongDelay (d) VLT Recognition

Study % Study %
ID ES (95% CI) Weight ID ES (95% CI) Weight
Bora et al 2007 0.50 (0.16, 0.84) 6.65
Bora et al 2007 0.45 (0.07, 0.82) 7.95
Cavanagh et al 2002 0.91 (0.32, 1.49) 2.25
Cavanagh et al 2002 0.61 (0.02, 1.20) 3.19
Clark et al 2002 0.28 (–0.18, 0.75) 3.59
Cubukcuoglu & Aydemir 0.68 (0.35, 1.01) 7.22 Clark et al 2002 0.36 (–0.10, 0.82) 5.26
Fleck et al 2003 0.66 (0.02, 1.30) 1.88 Cubukcuoglu & Aydemir 0.36 (–0.01, 0.73) 8.21
Goswami et al 2006 0.49 (0.07, 0.92) 4.24 Fleck et al 2003 –0.39 (–1.08, 0.31) 2.35
Hellvin et al 0.17 (–0.12, 0.47) 8.95 Kaya et al 2007 0.62 (0.14, 1.11) 4.84
Kaya et al 2007 1.16 (0.72, 1.60) 3.98 Kieseppa et al 2005 0.31 (–0.13, 0.74) 6.00
Kieseppa et al 2005 0.64 (0.24, 1.03) 4.89
Martinez-Aran et al 2004 0.55 (0.09, 1.01) 5.31
Martinez-Aran et al 2004 0.62 (0.21, 1.02) 4.70
Martinez-Aran et al 2007 0.50 (0.11, 0.90) 7.11
Martinez-Aran et al 2007 0.57 (0.21, 0.92) 6.15
Mur et al 2007 0.29 (–0.10, 0.69) 4.94 Mur et al 2007 0.26 (–0.19, 0.71) 5.59

Pfenning et al 0.02 (–0.53, 0.56) 2.59 Pfenning et al 0.32 (–0.34, 0.97) 2.64
Simonsen et al 2008 0.17 (–0.21, 0.56) 5.14 Simonsen et al 2008 0.25 (–0.15, 0.65) 6.98
Simonsen et al 2011 0.67 (0.35, 0.99) 7.58 Simonsen et al 2011 0.50 (0.16, 0.85) 9.57
Smith et al 2006 0.80 (0.27, 1.32) 2.83 Smith et al 2006 0.72 (0.21, 1.23) 4.34
Stoddart et al 2007 1.06 (0.51, 1.61) 2.57
Stoddart et al 2007 0.71 (0.13, 1.28) 3.37
Thompson et al 2005 0.52 (0.21, 0.84) 7.97
Thompson et al 2005 0.57 (0.25, 0.90) 10.64
Torrent et al 2006 0.71 (0.34, 1.09) 5.58
Torrent et al 2006 0.72 (0.31, 1.13) 6.65
Varma et al 0.70 (0.35, 1.06) 6.29
Overall (I-squared = 41.9%, p = 0.026) 0.55 (0.47, 0.64) 100.00 Overall (I-squared = 0.0%, p = 0.651) 0.46 (0.35, 0.57) 100.00

–1.61 0 1.61 –1.28 0 1.28

(e) VLT Recog–FP

Study %
ID ES (95% CI) Weight

Bora et al 2007 0.43 (0.06, 0.79) 10.13

Cavanagh et al 2002 0.68 (0.08, 1.28) 3.68

Cubukcuoglu & Aydemir 0.36 (0.01, 0.71) 10.70

Fleck et al 2003 0.09 (–0.57, 0.75) 3.03

Hellvin et al 0.14 (–0.16, 0.44) 15.00

Kieseppa et al 2005 0.46 (0.04, 0.87) 7.79

Mur et al 2007 0.51 (0.09, 0.93) 7.50

Pfenning et al –0.04 (–0.67, 0.59) 3.36

Simonsen et al 2008 0.11 (–0.29, 0.50) 8.58

Simonsen et al 2011 0.48 (0.15, 0.81) 12.28

Smith et al 2006 0.89 (0.37, 1.40) 5.05

Thompson et al 2005 0.49 (0.17, 0.82) 12.89

Overall (I-squared = 14.6%, p = 0.301) 0.38 (0.26, 0.50) 100.00

–1.4 0 1.4

Fig. 1. Forest plots showing the main effect of group (accounting for effect of age, IQ and gender) for the five outcome variables
associated with Verbal Learning Task (VLT).

8
 Euthymic bipolar cognition: IPDMA

(a) TMTA (b) TMTB

Study % Study %
ID ES (95% CI) Weight ID ES (95% CI) Weight

Balanza-Martinez et al 2005 –0.41 (–1.11, 0.29) 1.69 Balanza-Martinez et al 2005 –0.71 (–1.39, –0.03) 1.57

Bora et al 2007 –0.56 (–0.90, –0.21) 6.89 Bora et al 2007 –0.69 (–1.02, –0.36) 6.78

Cubukcuoglu & Aydemir –0.08 (–0.47, 0.32) 5.45 Cubukcuoglu & Aydemir –0.32 (–0.69, 0.04) 5.41
Dias et al 2009 –0.56 (–0.90, –0.23) 6.62
Dias et al 2009 –0.42 (–0.76, –0.09) 7.30
Dittman et al 2007 –0.30 (–0.59, –0.01) 8.63
Dittman et al 2007 –0.26 (–0.58, 0.05) 8.33
El-Badri et al 2001 –0.60 (–1.12, –0.09) 2.71
Goswami et al 2006 –0.49 (–0.89, –0.09) 5.14
Goswami et al 2006 –1.42 (–1.69, –1.15) 9.65
Macritchie et al –0.05 (–0.59, 0.49) 2.85
Macritchie et al –0.54 (–0.96, –0.12) 4.05
Martinez-Aran et al 2004 –0.84 (–1.22, –0.47) 5.92
Martinez-Aran et al 2004 –0.49 (–0.90, –0.07) 4.13
Martinez-Aran et al 2007 –0.68 (–1.03, –0.32) 6.66
Martinez-Aran et al 2007 –0.43 (–0.79, –0.06) 5.53
Mur et al 2007 –0.44 (–0.77, –0.12) 7.83
Mur et al 2007 –0.49 (–0.79, –0.19) 8.08
Smith et al 2006 –0.17 (–0.73, 0.39) 2.66
Smith et al 2006 –1.27 (–1.71, –0.83) 3.77
Soeiro-de-Souza & Soares-Bio –0.37 (–0.75, 0.02) 5.56
Soeiro-de-Souza & Soares-Bio –0.42 (–0.83, –0.01) 4.33
Stoddart et al 2007 –0.52 (–1.02, –0.02) 3.32
Stoddart et al 2007 –0.67 (–1.16, –0.18) 3.05
Szoke et al 2006 –0.68 (–1.00, –0.36) 8.12
Szoke et al 2006 –0.64 (–0.98, –0.31) 6.47
Thompson et al 2005 –0.50 (–0.82, –0.18) 8.02 Thompson et al 2005 –0.32 (–0.64, 0.01) 6.95
Torrent et al 2006 –0.71 (–1.10, –0.33) 5.62 Torrent et al 2006 –0.50 (–0.91, –0.09) 4.33
Varma et al –0.61 (–0.95, –0.27) 7.30 Varma et al –0.82 (–1.15, –0.49) 6.62
Zalla et al 2004 –0.46 (–1.24, 0.33) 1.34 Zalla et al 2004 –0.78 (–1.52, –0.04) 1.33
Overall (I-squared = 8.4%, p = 0.355) –0.49 (–0.58, –0.40) 100.00 Overall (I-squared = 68.6%, p = 0.000) –0.63 (–0.72, –0.55) 100.00

–1.24 0 1.24 –1.71 0 1.71

Fig. 2. Forest plots showing the main effect of group (accounting for effect of age, IQ and gender) for the two outcome variables
associated with Trail Making Test (TMTA and TMTB).

(a) WCSTCats. (b) WCSTPersev

Study % Study %
ID ES (95% CI) Weight ID ES (95% CI) Weight

Balanza-Martinez et al 2005 0.71 (0.10, 1.32) 3.33 Balanza-Martinez et al 2005 –0.72 (–1.27, –0.17) 4.04
Bora et al 2007 0.54 (0.15, 0.93) 8.25 Bora et al 2007 –0.47 (–0.86, –0.08) 8.02
Cubukcuoglu & Aydemir 0.40 (0.01, 0.78) 8.42 Cubukcuoglu & Aydemir –0.23 (–0.62, 0.17) 8.02
Fleck et al 2008 0.17 (–0.32, 0.67) 4.96 Fleck et al 2008 –0.18 (–0.69, 0.34) 4.61
Frangou et al 2005 0.64 (0.19, 1.09) 5.95 Frangou et al 2005 –0.20 (–0.69, 0.29) 5.22
Kieseppa et al 2005 0.12 (–0.33, 0.57) 6.11 Kieseppa et al 2005 –0.31 (–0.75, 0.13) 6.28
Martinez-Aran et al 2004 0.16 (–0.29, 0.62) 5.90 Martinez-Aran et al 2004 –0.51 (–0.95, –0.06) 6.23
Martinez-Aran et al 2007 0.07 (–0.30, 0.44) 8.97 Martinez-Aran et al 2007 –0.41 (–0.77, –0.04) 9.28
Melle et al 0.17 (–0.30, 0.63) 5.75 Melle et al 0.04 (–0.41, 0.50) 6.02
Mur et al 2007 0.24 (–0.17, 0.66) 7.19 Mur et al 2007 –0.27 (–0.69, 0.15) 6.94
Senturk et al 2007 0.36 (–0.13, 0.85) 5.12 Senturk et al 2007 –0.48 (–0.95, –0.01) 5.62
Simonsen et al 2011 –0.16 (–0.86, 0.54) 2.50 Simonsen et al 2011 0.10 (–0.62, 0.82) 2.38
Soeiro-de-Souza & Soares-Bio –0.18 (–0.57, 0.20) 8.34 Soeiro-de-Souza & Soares-Bio 0.52 (0.12, 0.92) 7.71
Szoke et al 2006 0.37 (0.03, 0.72) 10.22 Szoke et al 2006 –0.42 (–0.76, –0.08) 10.48
Torrent et al 2006 0.13 (–0.30, 0.56) 6.68 Torrent et al 2006 –0.48 (–0.90, –0.06) 7.01
Zalla et al 2004 0.51 (–0.22, 1.25) 2.29 Zalla et al 2004 –0.70 (–1.46, 0.06) 2.13
Overall (I-squared = 11.7%, p = 0.319) 0.26 (0.15, 0.37) 100.00 Overall (I-squared = 44.7%, p = 0.028) –0.29 (–0.40, –0.17) 100.00

–1.32 0 1.32 –1.46 0 1.46

Fig. 3. Forest plots showing the main effect of group (accounting for effect of age, IQ and gender) for the two outcome variables
associated with Wisconsin Card Sorting Task (WCSTCats. and WCSTPersev.).

illness variables that affected cognitive perfor- Table 5. Clinical indices of the patient group
mance, TMTA was affected by all four. Patients
M (SD) Range

Discussion Age at onset (n = 1129) 25.0 (8.7) 6–60

Illness duration (n = 1104) 13.8 (9.9) 0–51
This analysis of individual patient data across the No. of depressive episodes (n = 992) 5.6 (10.7) 0–100
31 studies provides further evidence that euthymic No. of manic episodes (n = 989) 3.4 (4.5) 0–88
bipolar patients exhibit moderate cognitive impair- Total no. of episodes (n = 1115) 11.6 (19.8) 0–200
No. of depressive hospitals (n = 271) 0.6 (1.4) 0–10
ments on a range of standard neuropsychological
No. of manic hospitals (n = 271) 1.4 (2.3) 0–15
tests. Cognitive deficits remain significant even Total no. of hospitalisations (n = 806) 2.9 (3.8) 0–40
after controlling for key baseline factors such as
age, IQ and gender that are known to affect neuro-
psychological test performance. The current level
of minor depressive symptoms and the effects of with bipolar disorder over and above the known
some drug treatments may contribute to these confounding factors.
effects but cannot explain them. Thus, there is sig- The effect sizes for such deficits were lower
nificant residual cognitive impairment associated (0.26–0.63) than those reported in prior meta-anal-

9
Bourne et al.

yses (6–8, 10) (ds = 0.5–1.0). This reduction in more standardized computerized formats locally
observed effect sizes is in part due to controlling or even on line.
better for the effect of age, IQ and gender. How- Nevertheless, the group effect sizes allow confi-
ever, we were also able to include unpublished dence that a substantial average effect is present
studies which often had the lowest effect sizes [e.g. for the domains of attention/working memory,
Hellvin et al. (50) and A. Pfennig, M. Alda, T. verbal memory, speed and executive function. It is
Young, et al. (manuscript in preparation) for VLT somewhat easier to say what cannot explain these
Total1–5, LongDelay and Recog-FP; Cubukcuo- effects, than to say what can. Residual mood
glu & Aydemir (49) and A. Macritche (manuscript symptoms within the patient group were under-
in preparation) for TMTA and TMTB; A. Varma standably confounded with group. However, our
(manuscript in preparation) for FDS and RDS; analysis suggests that residual symptom scores in
M.G. Soeiro-de-Souza & D. Soares-Bio (manu- the patient group cannot explain much of the dif-
script in preparation) for WCSTCats.; and Hellvin ference found between the groups across the vari-
et al. (50) and M.G. Soeiro-de-Souza & D. Soares- ous tests. Cognitive deficits are also not simply
Bio (manuscript in preparation) for WCSTPer- explained as side-effects of drug therapy. This has
sev.]. This suggests the field has had some previously been the subject of debate; some studies
impact from publication bias, which perhaps is suggesting that antipsychotic drugs may cause
unsurprising. some cognitive impairment (62, 63) and others sug-
Specifically, the following effect sizes were found gesting no drug effect on cognitive performance
(compared to prior studies) in the following cogni- (64). The present analysis suggests that most neu-
tive domains: i) verbal memory – Total Score effect ropsychological tests do not exhibit any significant
size = 0.51 (prior studies = 0.90–0.81), Short effect attributable to drug treatment. The only pos-
Delay effect size = 0.48 (prior studies = 0.85–0.73), sible exception is on measures of verbal memory
Long Delay effect size = 0.55 (prior stud- with antipsychotics having an impairing effect on
ies = 0.85–0.71), Recognition effect size = 0.46 VLT Total1–5 and drug-free status being associ-
(prior study = 0.43), Recog-FP effect size = 0.38; ated with improved performance on VLT Total1–5
ii) visual scanning speed – TMTA effect size = 0.49 and LongDelay (relative to any drug). However,
(prior studies = 0.82–0.60); iii) working memory any potential implied drug effects must be treated
capacity – FDS effect size = 0.30 (prior stud- with caution due to the potential for confounding
ies = 0.47–0.37); iv) executive function – TMTB by indication. For example, a history of psychosis
effect size = 0.63 (prior studies = 0.99–0.55), RDS may be related to specific working memory impair-
effect size = 0.60 (prior studies = 1.02–0.54), ments (65–67), and those with a history of psycho-
WCSTCats. effect size = 0.26 (prior stud- sis are also likely to be those currently taking
ies = 0.69–0.52); v) response inhibition = WCST- antipsychotics (68). We could not analyse the effect
Persev. = 0.29 (prior studies = 0.88–0.70). of polypharmacy, which is common in clinical
The high heterogeneity of some tests appears samples, but not in these research samples. It is
to underlie the differences in the results of prior likely that there was a deliberate effort to exclude
meta-analyses. The variation in effect sizes symptomatic and heavily medicated patients from
between the previously published meta-analysis these studies given the intention was usually to
(Table 1) is likely to have been due to variations reduce the confounds between the patient and
in the studies included. In turn, the range of control groups.
effect sizes produced by including a different sub- If illness course had had a negative impact on
set of studies can be directly explained by the rel- cognition, it would potentially be a key finding; it
atively high level of heterogeneity revealed in this could imply that neuropsychological outcome
sample by our analysis (typically 39–84%; see measures are sensitive to treatment. In a partial
Table 4) especially for some tests. The test with support of this hypothesis, some of the neuropsy-
the most heterogeneity in this analysis was chological measures correlated with illness inten-
TMTB. TMTB is known to have considerable sity variables, for example number of manic
variability across test sites (61), thus there episodes appears to affect performance on certain
appears to be a strong case for trying to refine VLT measures, whilst TMTA appears to be espe-
the operationalization of TMTB as well as VLT cially sensitive to potential illness progression
(encoding and short term recall) and Digit Span effects. However, the magnitude of these associa-
(Forward and Reverse). Each test taps domains tions may be unreliable for various reasons. First,
of function markedly impaired in bipolar patients the impact of illness may not be simply cumulative,
as shown by the large average effect sizes. One and the largest effects may occur early in the illness
important possibility would be to present them in course, as appears likely in schizophrenia (69).

10
 Euthymic bipolar cognition: IPDMA

Second, measures of illness severity that depend on indicate that deficits are stable despite long-term
counting episodes in mature samples of patients lithium therapy (76, 77).
are of uncertain validity. Quantifying depressive As with all analyses of neuropsychological per-
episodes when so much of the depressive burden of formance, this study’s findings and conclusions are
bipolar disorder is chronic, subsyndromal and limited by the reliability, validity and psychometric
poorly recalled is questionable; indeed, we found properties of the individual neuropsychological
no associations with number of depressive epi- tests. The high levels of heterogeneity found in this
sodes. Positive findings for more memorable study and the previous standard meta-analyses
events, like manic episodes and numbers of hospi- (6–9, 12, 13) for some measures highlight the need
talizations, appear more likely to be valid and did for standardization in test presentation to try and
produce some significant results in this analysis. meet this limitation. Indeed, the high levels of heter-
The hypothesis that much of the apparent cogni- ogeneity consistently found for some measures
tive impairment of bipolar disorder is attributable raises the question as to whether it is meaningful to
to the accumulated impact of the illness course combine them in a meta-analysis at all. This study
remains plausible but not proven by the present is also limited by the response bias of authors allow-
study. Only adequately powered prospective ing access to their primary data sets. Furthermore,
studies in early stages of illness will establish the it is acknowledged that this study considered out-
effect beyond doubt. come measures from a relatively small number of
Although the range in effect sizes reported here neuropsychological tests. However, despite being
appears to support previous suggestions that exec- limited to those primary studies that consented to
utive function and memory may be especially provide data, and partly because the analysis was
affected in bipolar disorder (6, 8, 9), it is also nota- limited to the most frequently used neuropsycho-
ble that all of the effect sizes reported here could be logical tests, this study contained sample sizes sub-
considered to be small to medium (70) in magni- stantially greater than many of the prior standard
tude across all the cognitive domains investigated. meta-analyses and thus represents a major data
Our results could therefore also be interpreted as synthesis. Furthermore, by using IPDMA (rather
being consistent with the notion of cognitive than standard meta-analysis) this study was both
impairment in bipolar disorder being a relatively able to i) provide the least confounded estimates of
non-specific effect on multiple functional brain net- the effect size relating to cognitive impairment in
works. This can be related to similarly non-specific euthymic bipolar patients and ii) provide the first
imaging findings suggesting lateral ventricle analysis of potential medication and illness severity
enlargement (effect size = 0.39) and increased rates effects on neuropsychological performance in a
of deep white matter hyperintensities without grey statistically valuable sample size.
matter volume decrements (71) in the many imag- In summary, this reanalysis provides further
ing studies conducted in bipolar patients. evidence that euthymic bipolar patients exhibit
Although these structural abnormalities can be significant cognitive impairment on a range of
greater in older patients they are also found in neuropsychological tests. These impairments
samples of similar mean age as the sample in this remain substantial but less than previous work
study (71). The evolving evidence for widely dis- (including previous meta-analyses) has suggested
tributed disturbances in white matter structure (1–4, 6–10). The advantage of IPDMA in control-
from diffusion tensor imaging is also supportive of ling for a greater range of confounding factors
an underlying functional neuropathology (72). and the inclusion of unpublished studies accounts
Although its aetiology remains poorly understood, for this. The impairment effect appears largely
a contribution from intracellular mechanisms regu- independent of drug treatment. Performance on
lating oxidative stress is one hypothesis that is some neuropsychological tests appears to have
assuming increasing importance (73). Given the deteriorated further as illness progressed (i.e.
putative neuroprotective effects of lithium (74, 75), number of episodes increased) but longitudinal
an improved cognitive performance for those data from earlier in the illness course are needed
patients taking lithium relative to those lithium to show that the relationship is causal and clini-
free might have been expected. However, no such cally important. Finally, this review and reanalysis
effect was found; either because lithium does not has highlighted the variability and heterogeneity
enhance cognitive performance or because any between individual primary studies. This means
neuroprotective effect is dependent upon factors, the field remains polarized between the certainty
such as chronic use, which could not be estimated that cognitive impairment is a feature of bipolar
in this dataset. In support of the former ‘ineffective disorder and uncertainty, for example about its
hypothesis’, two recent longitudinal cohort studies heritability, specificity or the impact of illness

11
Bourne et al.

intensity. Specific and correct findings on the latter from AstraZeneca. Dr. Stoddart currently works for a consul-
may be reasonably based on studies that are well tancy firm that has pharmaceutical companies among its cli-
ents. Dr. Vieta has received grants and served as consultant,
conducted but too small for confidence and too advisor or CME speaker for the following entities: Adamed,
subtle to be replicated in cohorts of convenience. Alexza, Almirall, AstraZeneca, Bial, Bristol-Myers Squibb,
On the other hand, small studies can always gener- Elan, Eli Lilly, Ferrer, Forest Research Institute, Gedeon
ate false positives findings, and this is too often Richter, Glaxo-Smith-Kline, Janssen-Cilag, Jazz, Johnson &
forgotten in the field (78). The present result, from Johnson, Lundbeck, Merck, Novartis, Organon, Otsuka,
Pfizer, Pierre-Fabre, Qualigen, Roche, Sanofi-Aventis, Servier,
a study sample larger than the samples reported in Shering-Plough, Shire, Solvay, Sunovion, Takeda, Teva, the
three of the previous meta-analyses of published Spanish Ministry of Science and Innovation (CIBERSAM),
data sets, may well be giving us the true picture. A the Seventh European Framework Programme (ENBREC),
clear goal for future research is operationally to the Stanley Medical Research Institute, United Biosource
refine all test procedures and variables being mea- Corporation and Wyeth.
sured to reduce heterogeneity and combine data
prospectively across centres to obtain the necessary References
power essential to statistical confidence. 1. Bearden CE, Hoffman KM, Cannon TD. The neuropsy-
chology and neuroanatomy of bipolar affective disorder:
A critical review. Bipolar Disord 2001;3:106–150.
Acknowledgements 2. Quraishi S, Frangou S. Neurospychology of bipolar disor-
der. J Affect Disord 2002;72:209–226.
This paper was partially supported by a Seventh Framework
3. Savitz J, Solms M, Ramesar RS. Neuropsychological dys-
Programme grant from the European Union to the European
function in bipolar affective disorder: a critical opinion.
Network of Bipolar Research Expert Centres (ENBREC),
Bipolar Disord 2005;7:216–235.
Grant No. Health-F2-2009-223102.
4. Martinez-Aran A, Vieta E, Reinares M et al. Cognitive
function across manic or hypomanic, depressed, and
Declaration of interest euthymic states in bipolar disorder. Am J Psychiatry
2004;161:262–270.
Drs. Bora, Bourne, Craddock, Cubukcuoglu, Dittmann, Fleck, 5. Goodwin GM, Anderson I, Arango C et al. ECNP consen-
Gallagher, Geddes, Jones, Kiesepp€ a, Leboyer, Martınez-Aran, sus meeting. Bipolar depression. Nice, March 2007. Eur
Melle, Moore, Mur, Raust, Rogers, Senturk, Simonsen, So- Neuropsychopharmacol 2008;18:535–549.
ares-Bio, Smith, Soeiro-de-Souza, Sundet, Sz€ oke, Thompson, 6. Arts B, Jabben N, Krabbendam L, van Os J. Meta-analyses
Torrent, Tzagarakis, Worhunsky and Zalla declare that they of cognitive functioning in euthymic bipolar patients and
have no conflicts of interest over the past 2 years. Dr. Andreas- their first-degree relatives. Psychol Med 2007;38:771–785.
sen has received speakers’s honorarium from Lilly, Lundbeck 7. Bora E, Yucel M, Pantelis C. Cognitive endophenotypes
and GSK. Dr. Clark is a consultant for Cambridge Cognition of bipolar disorder: a meta-analysis of neuropsychological
Ltd. Dr. Aydemir has participated in a clinical trial sponsored deficits in euthymic patients and their first-degree relatives.
by AstraZeneca, received speaker honoraria from Lundbeck, J Affect Disord 2009;113:1–20.
AstraZeneca, Janssen-Cilag and Pfizer and consultant for Ser- 8. Robinson LJ, Thompson JM, Gallagher P et al. A meta-
vier. Dr Balanza-Martınez has received grants and served as analysis of cognitive deficits in euthymic patients with
consultant, advisor or CME speaker from Angelini, AstraZen- bipolar disorder. J Affect Disord 2006;93:105–115.
eca, Bristol-Myers-Squibb, Grunenthal, Janssen, Juste, the 9. Torres IJ, Boudreau VG, Yatham LN. Neuropsychological
Spanish Ministry of Science and Innovation (CIBERSAM) functioning in euthymic bipolar disorder: a meta-analysis.
and ′Fundaci on Alicia Koplowitz′. Dr. Bauer has received Acta Psychiatr Scand 2007;116:17–26.
grant/research support from The Stanley Medical Research 10. Robinson LJ, Ferrier IN. Evolution of cognitive impair-
Institute, NARSAD, Deutsche Forschungsgemeinschaft and ment in bipolar disorder: a systematic review of cross-
the European Commission (FP7). He is a consultant for sectional evidence. Bipolar Disord 2006;8:103–116.
Alkermes, AstraZeneca, BristolMyers Squibb, Ferrer Internac- 11. Krabbendam L, Arts B, van Os J, Aleman A. Cognitive
ional, Janssen, Lilly, Lundbeck, Otsuka, Servier, Takeda. Dr. functioning in patients with schizophrenia and bipolar dis-
Bauer has received speaker honoraria from AstraZeneca, order: a quantitative review. Schizophr Res 2005;80:
BristolMyers Squibb, GlaxoSmithKline, Lilly, Lundbeck, 137–149.
Otsuka. Pfizer. Dr. Brissos has been working full time as 12. Mann-Wrobel MC, Carreno JT, Dickinson D. Meta-analy-
Medical Affairs Manager for Janssen Pharmaceutical. Dr. sis of neuropsychological functioning in euthymic bipolar
Cavanagh has received investigator-originated research grant disorder: an update and investigation of moderator vari-
funding from Pfizer and Biogen IDEC. Dr. Dias is consultant ables. Bipolar Disord 2011;13:334–342.
for Angelini Pharmaceutical, Portugal and has received educa- 13. Kurtz MM, Gerraty RT. A meta-analytic investigation of
tional grants from Lundbeck, Sanofi-Aventis, AstraZeneca neurocognitive deficits in bipolar illness: profile and effects
and Bristol-Myers Squibb. Dr. Ferrier has received speaker of clinical state. Neuropsychology 2009;5:551–562.
honoraria for lectures given at educational meetings sponsored 14. The Psychiatric GWAS Consortium Steering Committee.
by Astra Zeneca and Organon. Dr. Frangou has participated A framework for interpreting genome-wide association
in advisory boards for Janssen-Cilag and Ferrer Grupo studies of psychiatric disorders. Mol Psychiatry 2009;14:
and has been a speaker for Janssen-Cilag. Dr. Goodwin 10–17.
has received grants/research support, consulting fees and 15. Rivas MA, Beaudoin M, Gardet A et al. Deep resequencing
honoraria from AstraZeneca, Bristol-Myers Squibb, Eisai, of GWAS loci identifies independent rare variants associ-
Eli Lilly, Lundbeck, P1Vital, Servier, Takeda and Teva. Dr. ated with inflammatory bowel disease. Nat Genet
Pfennig has received research support and speaker honoraria 2011;43:1066–1075.

12
 Euthymic bipolar cognition: IPDMA

16. Belmonte Mahon P, Pirooznia M, Goes FS et al. Genome- 36. Kieseppa T, Tuulio-Henriksson A, Haukka J et al. Memory
wide association analysis of age at onset and psychotic and verbal learning functions in twins with bipolar-I disor-
symptoms in bipolar disorder. Am J Med Genet B Neuro- der and the role of information-processing speed. Psychol
psychiatr Genet 2011;156:370–378. Med 2005;35:205–215.
17. Clarke MJ. Individual patient data meta-analyses. Best 37. Martinez-Aran A, Vieta E, Torrent C et al. Functional
Pract Res Clin Obstet Gynaecol 2005;19:47–55. outcome in bipolar disorder: the role of clinical and cogni-
18. Delis DC, Kramer JH, Kaplan E, Ober BA. California ver- tive factors. Bipolar Disord 2007;9:103–113.
bal learning test: adult version. San Antonio, TX: The 38. Mur M, Portella MJ, Martinez-Aran A, Pifarre J, Vieta
Psychological Corporation, 1987. E. Persistant neuropsychological deficit in euthymic bipo-
19. Rey A. L’examen clinique en psychologie. Paris: Presses lar patients: executive function as a core deficit. J Clin Psy-
Universitaires de France, 1964. chiatry 2007;68:1078–1086.
20. Reitan RM. Validity of trail making test as an indication 39. Senturk V, Goker C, Bilgic A et al. Impaired verbal mem-
of organic brain disease. Percept Mot Skills 1958;8: ory and otherwise spared cognition in remitted bipolar
271–276. patients on monotherapy with lithium or valproate. Bipo-
21. Wechsler D. Wechsler adult intelligence scale – revised: lar Disord 2007;9:136–144.
manual. New York: Psychological Corporation, 1981. 40. Smith DJ, Muir WJ, Blackwood DHR. Neurocognitive
22. Heaton RK. Wisconsin card sorting test manual. Odessa, impairment in euthymic young adults with bipolar spec-
FL: Psychological Assessment Resources Inc., 1981. trum disorder and recurrent major depressive disorder.
23. Yatham LN, Torres IJ, Malhi GS et al. The International Bipolar Disord 2006;8:40–46.
Society for Bipolar Disorders – battery for assessment of 41. Stoddart SDR, Craddock NJ, Jones LA. Differentiation of
neurocognition (ISBD-BANC). Bipolar Disord 2010;12: executive function and attention impairments in affective
351–363. illness. Psychol Med 2007;37:1613–1623.
24. Moher D, Liberati A, Tetzlaff J, Altman DG, Group TP. 42. Szoke A, Schurhoff F, Golmard J et al. Familial resem-
Preferred reporting items for systematic reviews and meta- blance for executive functions in families of schizo-
analyses: the PRISMA statement. J Clin Epidemiol phrenic and bipolar patients. Psychiatry Res 2006;144:
2009;62:1006–1012. 131–138.
25. Simonsen C, Sundet K, Vaskinn A et al. Neurocognitive 43. Thompson JM, Gallagher P, Hughes JH et al. Neurocogni-
profiles in bipolar I and bipolar II disorder: differences in tive impairment in euthymic patients with bipolar affective
pattern and magnitude of dysfunction. Bipolar Disord disorder. Br J Psychiatry 2005;186:32–40.
2008;10:245–255. 44. Torrent C, Martinez-Aran A, Daban C et al. Cognitive
26. Balanza-Martinez V, Tabares-Seisdedos R, Selva-Vera G impairment in bipolar II disorder. Br J Psychiatry
et al. Persistent cognitive dysfunctions in bipolar I disor- 2006;189:254–259.
der and schizophrenic patients: a 3-year follow-up study. 45. Zalla T, Joyce C, Szoke A et al. Executive dysfunctions
Psychother Psychosom 2005;74:113–119. as potential markers of familial vulnerability to bipolar
27. Bora E, Vahip S, Akdeniz F et al. The effect of previous disorder and schizophrenia. Psychiatry Res 2004;121:
psychotic mood episodes on cognitive impairment in 207–217.
euthymic bipolar patients. Bipolar Disord 2007;9:468–477. 46. Dias VV, Brissos S, Frey BN, Andreazza AC, Cardoso C,
28. Cavanagh JTO, van Beck M, Muir M, Blackwood DHR. Kapczinski F. Cognitive function and serum levels of brain-
Case-control study of neurocognitive function in euthymic derived neurotrophic factor in patients with bipolar disor-
patients with bipolar disorder: an association with mania. der. Bipolar Disord 2009;11:663–671.
Br J Psychiatry 2002;180:320–326. 47. Simonsen C, Sundet K, Vaskinn A et al. Neurocognitive
29. Clark L, Iversen SD, Goodwin GM. Sustained attention dysfunction in bipolar and schizophrenia spectrum disor-
deficit in bipolar disorder. Br J Psychiatry 2002;180: 313–319. ders depends on history of psychosis rather than diagnos-
30. Dittmann S, Seemuller F, Schwarz MJ et al. Association of tic group. Schizophr Bull 2011;37:73–83.
cognitive deficits with elevated homocysteine levels in eu- 48. Kaya E, Aydemir O, Selcuki D. Residual symptoms in
thymic bipolar patients and its impact on psychosocial bipolar disorder: the effect of the last episode after remis-
functioning: preliminary results. Bipolar Disord sion. Prog Neuropsychopharmacol Biol Psychiatry
2007;9:63–70. 2007;31:1387–1392.
31. El-Badri SM, Ashton CH, Moore PB, Marsh VR, Ferrier 49. Cubukcuoglu Z, Aydemir O. Specifity and sensitivity of
IN. Electrophysiological and cognitive function in young neurocognitive tests in bipolar disorder: A comparative
euthymic patients with bipolar affective disorder. Bipolar study. Unpublished thesis, 2011.
Disord 2001;3:79–87. 50. Hellvin T, Sundet K, Simonsen C et al. Neurocognitive
32. Fleck DE, Shear PK, Zimmerman ME et al. Verbal memory functioning in patients recently diagnosed with bipolar
in mania: effects of clinical state and task requirements. disorder. Bipolar Disord 2012;14:227–238.
Bipolar Disord 2003;5:375–380. 51. Hamilton M. A rating scale for depression. J Neurol
33. Fleck DE, Shear PK, Madore M, Strakowski SM. Wiscon- 1960;23:55–62.
sin card sorting test performance in bipolar disorder: 52. Montgomery SA, Asberg M. A new depression scale
effects of mood state and early course. Bipolar Disord designed to be sensitive to change. Br J Psychiatry
2008;10:539–545. 1979;134:382–389.
34. Frangou S, Donaldson S, Hadjulis M, Landau S, Goldstein 53. Rush AJ, Carmody T, Reimitz PE. The inventory of depres-
LH. The Maudsley bipolar disorder project: executive dys- sive sympotmatology (IDS): clinician (IDS-C) and self-
function in bipolar I disorder and its clinical correlates. report (IDS-SR) ratings of depressive symptoms. Int J
Biol Psychiatry 2005;58:859–864. Methods Psychiatr Res 2000;9:45–59.
35. Goswami U, Sharma A, Khastigir U et al. Neuropsycholog- 54. Young RC, Briggs JT, Ziegler VE, Meyer DA. A rating
ical dysfunction, soft neurological signs and social disabil- scale for mania: reliability, validity, and sensitivity. Br J
ity in euthymic patients with bipolar disorder. Br J Psychiatry 1978;133:429–435.
Psychiatry 2006;188:366–373.

13
Bourne et al.

55. Altman EG, Hedeker DR, Janicak PG, Peterson JL, Davis 68. Torrent C, Martinez-Ar an A, Daban C et al. Effects of
JM. The clinician-administered rating scale for mania atypical antipsychotics on neurocognition in euthymic
(CARS-M): development, reliability, and validity. Biol bipolar patients. Compr Psychiatry 2011;52:203–213.
Psychiatry 1994;36:124–134. 69. Mohamed S, Paulsen JS, O’leary D, Arndt S, Andreasen N.
56. Bech P, Boewig TG, Dein E, Jacobsen O, Gram LF. Quanti- Generalized cognitive deficits in schizophrenia: a study
tative rating of manic states. Acta Psychiatr Scand of first-episode patients. Arch Gen Psychiatry 1999;56:
1975;59:1–6. 749–754.
57. Nelson HE, Willison J. National adult reading test man- 70. Cohen JW. Statistical power analysis for the behavioural
ual. 2nd ed. Windsor: NFER-Nelson, 1991. sciences. 2nd ed. Hillsdale NJ: Lawrence Erlbaum
58. Wechsler D. Wechsler abbreviated scale of intelligence Associates, 1988.
(WASI). San Antonio, TX: Harcourt Assessment, 1999. 71. Kempton MJ, Geddes J, Ettinger U, Williams SCR, Grasby
59. Wechsler D. Wechsler memory scale – revised. New York: PM. Meta-analysis, database, and meta-regression of 98
Psychological Corporation, 1987. structural imaging studies in bipolar disorder. Arch Gen
60. Higgins JPT, Thompson SG. Quantifying heterogeneity in a Psychiatry 2008;65:1017–1032.
meta-analysis. Stat Med 2002;21:1539–1558. 72. Vederine FE, Wessa M, Leboyer M, Houenou J. A meta-
61. Nuechterlein KH, Green MF, Kern RS et al. The analysis of whole-brain diffusion tensor imaging studies in
MATRICS consensus cognitive battery, part 1: test selec- bipolar disorder. Prog Neuropsychopharmacol Biol Psy-
tion, reliability, and validity. Am J Psychiatry 2008;165: chiatry 2011;35:1820–1826.
203–213. 73. Andreazza AC, Kauer-Sant’anna M, Frey BN et al. Oxida-
62. Donaldson S, Goldstein LH, Landau S, Raymont V, Fran- tive stress markers in bipolar disorder: a meta-analysis. J
gou S. The Maudsley Bipolar Disorder Project: the effect Affect Disord 2008;111:135–144.
of medication, family history, and duration of illness on 74. Shao L, Young LT, Wang J. Chronic treatment with mood
IQ and memory in bipolar I disorder. J Clin Psychiatry stabilizers lithium and valproate prevents excitotoxicity by
2003;64:86–93. inhibiting oxidative stress in rat cerebral cortical cells. Biol
63. Jamrozinski K, Gruber O, Kemmer C, Falkai P, Scherk H. Psychiatry 2005;58:879–884.
Neurocognitive functions in euthymic bipolar patients. 75. Tan H, Young LT, Shao L, Che Y, Honer WG, Wang J.
Acta Psychiatr Scand 2009;119:365–374. Mood stabilizer lithium inhibits amphetamine-increased
64. Goswami U, Sharma A, Varma A et al. The neurocogni- 4-hydroxynonenal-protein adducts in rat frontal cortex.
tive performance of drug-free and medicated euthymic Int J Neuropsychopharmacol 2012;15:1275–1285.
patients do not differ. Acta Psychiatr Scand 2009;120: 76. Mora E, Portella MJ, Forcada I, Vieta E, Mur M. Persis-
456–463. tence of cognitive impairment and its negative impact on
65. Glahn DC, Bearden CE, Barguil M et al. The neurocogni- psychosocial functioning in lithium-treated, euthymic
tive signature of psychotic bipolar disorder. Biol Psychia- bipolar patients: a 6-year follow-up study. Psychol Med
try 2007;62:910–916. 2012; doi: 10.1017/S0033291712001948. [Epub ahead of
66. Glahn DC, Therman S, Manninen M et al. Spatial working print].
memory as an endophenotype for schizophrenia. Biol Psy- 77. Torrent C, Martinez-Aran A, del Mar Bonnin C et al.
chiatry 2003;53:624–626. Long-term outcome of cognitive impairment in bipolar
67. Glahn DC, Bearden CE, Cakira S et al. Differential work- disorder. J Clin Psychiatry 2012;73:e899–e905.
ing memory impairment in bipolar disorder and schizo- 78. Ioannidis JPA. Why most published research findings are
phrenia: effects of lifetime history of psychosis. Bipolar false. PLoS Med 2005;2:e124.
Disord 2006;8:117–123.

14