Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
Contents lists available at ScienceDirect
Progress in Neuro-Psychopharmacology & Biological
Psychiatry
journal homepage: www.elsevier.com/locate/pnp
Bipolar disorder and schizophrenia share a similar deficit in semantic
inhibition: A meta-analysis based on Hayling Sentence Completion
Test performance
Kui Wang a,1, Li-Ling Song a,1, Eric F.C. Cheung b, Simon S.Y. Lui a,b,c, David H.K. Shum d, Raymond C.K. Chan a,⁎
a
Neuropsychology and Applied Cognitive Neuroscience Laboratory, Key Laboratory of Mental Health, Institute of Psychology, Chinese Academy of Sciences, Beijing, China
Castle Peak Hospital, Hong Kong Special Administrative Region, China
c
University of Chinese Academy of Sciences, Beijing, China
d
Behavioural Basis of Health Program, Griffith Health Institute and School of Applied Psychology, Griffith University, Brisbane, Australia
b
a r t i c l e
i n f o
a b s t r a c t
Article history:
Received 15 March 2013
Received in revised form 12 July 2013
Accepted 12 July 2013
Available online 22 July 2013
Bipolar disorder (BD) is associated with deficits in executive function similar to that found in schizophrenia (SZ).
However, very few studies have examined whether a specific component of executive function, namely, semantic
inhibition, is differentially impaired in BD and SZ. The present study reports the results of a meta-analysis of performance on a theory-driven test of semantic inhibition, namely, the Hayling Sentence Completion Test (HSCT),
in patients with BD and SZ, and to examine differential group impairments. The Comprehensive Meta-Analysis
Software package was used to calculate the mean effect sizes for group differences on different measures
of HSCT. A total of 13 studies were included in the meta-analysis. Effect sizes for six HSCT measures were calculated. These included: Total Latency of Task A, Total Latency of Task B, Suppression Time, Total Error of Task B,
Type A Error of Task B, and Type B Error of Task B. When compared with healthy controls, medium-to-large effect
sizes were observed in both groups for each HSCT measure. Interestingly, the effect sizes for BD and SZ groups
were comparable. These results suggest that patients with SZ and patients with BD are impaired in both task initiation and task inhibition of executive function and these impairments are similar in magnitude for both
disorders.
© 2013 Elsevier Inc. All rights reserved.
Keywords:
Bipolar disorder
Hayling Sentence Completion Test
Inhibition
Meta-analysis
Schizophrenia
Semantic processing
Contents
1.
2.
Introduction . . . . . . . . .
Methods . . . . . . . . . . .
2.1.
Literature search . . . .
2.2.
Meta-analysis procedure
3.
Results . . . . . . . . . . .
3.1.
Total Latency of Task A .
3.2.
Total Latency of Task B .
3.3.
Suppression time . . .
3.4.
Total Error of Task B . .
3.5.
Type A Error in Task B .
3.6.
Type B Error in Task B .
4.
Discussion . . . . . . . . . .
Acknowledgments . . . . . . . . .
References . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
154
155
155
155
156
156
157
157
157
158
158
159
159
159
Abbreviations: BD, bipolar disorder; BDD, bipolar disorder in the depression phase; BDM, Bipolar disorder in the manic phase; BDR, bipolar disorder in the remitted phase; DSM,
Diagnostic and Statistical Manual of Mental Disorders; HSCT, Hayling Sentence Completion Test; TD, thought disorder; SZ, Schizophrenia.
⁎ Corresponding author at: 16 Lincui Road, Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China. Tel./fax: +86 10 64836274.
E-mail address: [email protected] (R.C.K. Chan).
1
These authors contribute equally to the writing up of this paper.
0278-5846/$ – see front matter © 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.pnpbp.2013.07.012
154
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
1. Introduction
Traditionally, while bipolar disorder (BD) has abnormal emotional
processes as its key feature, schizophrenia (SZ) is characterized by cognitive deficits. However, recent studies have found that patients with BD
also have deficits in various cognitive functions (Arts et al., 2008;
Krabbendam et al., 2005; Quraishi and Frangou, 2002). Considering
that the two disorders also share many similar symptoms, these recent
findings raised the question about whether the observed cognitive deficits in the two patient groups are more related to diagnosis or to pattern of symptoms.
Various definitions of executive functions have been put forward
(Burgess et al., 2000; Royall et al., 1993; Stoddart et al., 2007), with
the one proposed by Burgess et al. (2000) being most widely accepted.
According to them, executive function refers to a wide range of cognitive processes and behavioral abilities such as problem-solving, sequencing, verbal reasoning, the ability to sustain attention, planning,
resistance to interference, utilization of feedback, multitasking, cognitive flexibility and the ability to deal with novelty. Among the various
cognitive deficits identified, previous studies have highlighted executive dysfunctions in BD and SZ for several reasons. First, executive functions are closely associated with everyday functioning, and are probably
most impaired compared to other cognitive processes in individuals
with SZ (Frangou, 2010) and BD (Arts et al., 2008). Second, executive
dysfunctions are related to symptoms observed in BD and SZ. Executive
functions are primarily sub-served by the frontal lobes (Andreasen et al.,
1996) and frontal lobe lesions are associated with both the failure to
suppress inappropriate responses and the lack of responses (Shallice,
1988). While failure to suppress inappropriate responses is associated
with thought disorder and reality distortion (delusion and hallucination) in SZ and mania in BD, lack of responses is associated with psychomotor poverty in SZ and depression in BD (Kravariti et al., 2005). Third,
impaired executive functions may be a common trait marker in BD and
SZ (Breton et al., 2011, Frangou et al., 2005a, Morey et al., 2005).
In this study we focused on the inhibition function in individuals
with BD and SZ because factorial-analytic studies have repeatedly
shown that semantic inhibition is a very important factor in a battery
of executive function tests, in both healthy (Chan, 2001) and patient
populations (Chan et al., 2004). The Hayling Sentence Completion Test
(HSCT) is a test developed to assess inhibition (Burgess et al., 2000)
and it has been widely used in clinical practice. In the HSCT, individuals
are presented with incomplete sentences with the final word omitted
but is strongly suggested by the context. Individuals are asked to complete the sentence in either a logical (Task A, initiation section) or illogical manner (Task B, inhibition section). In Task B any word which is
semantically associated with the sentence should be avoided, thus test
takers have to inhibit a strongly cued and automatic response. For instance, responding with the word “ship” to the sentence “the captain
went down with the sinking __” is correct when undertaking Task A,
but incorrect when undertaking Task B (Type A error). Moreover,
words such as ‘airplane’, ‘bus’, ‘waterman’ which are semantically associated with the whole sentence context are also scored incorrect for
Task B (Type B error). Thus, in completing Task B, test takers are required not only to suppress a pre-potent response but also to plan and
manipulate information in working memory. Shorter latency and
fewer errors in Task A or Task B indicate better initiation or inhibition
function.
Significant correlations between HSCT scores and self-reported measures of attentional impulsivity have been established in remitted BD
patients, suggesting that poor response inhibition may be related to impulsivity in these patients (Christodoulou et al., 2006). Impaired HSCT
performance has been observed in patients with different symptomatology, including patients with BD (Stoddart et al., 2007). Moreover,
Dixon et al. (2004) studied the relationship between inhibition function
and symptomatology by recruiting 15 manic, 15 depressed, 15 remitted
patients with BD and 30 controls. Even with the modest number of
participants in each group, the authors found that each of the three BD
subgroups had longer latency, larger error rate in Task A, and decreased
use of strategy in Task B (e.g., reporting objects in the testing environment). Similarly, compromised HSCT performance in euthymic BD patients (de Almeida Rocca et al., 2008) and remitted BD patients
(Frangou et al., 2005b) have been reported. These findings, therefore,
suggest that impaired performance in HSCT may be an enduring feature
of BD and not a secondary deficit due to mood symptoms.
Nathaniel-James et al. (1996) first reported that patients with SZ had
difficulties in performing the HSCT when compared to healthy controls.
Patients with SZ showed longer response latency in Task A and more errors in Task B, indicating clear deficits in response initiation and inhibition. Since then, a large number of studies have repeatedly found that SZ
patients exhibit impaired performance on the HSCT, especially in Task B
where inhibition is needed (Chan et al., 2012; Chan and Chen, 2004;
Chan et al., 2004, 2010; Groom et al., 2008; Joshua et al., 2009;
Marczewski et al., 2001; Nathaniel-James et al., 1996, Royer et al.,
2009a; Waters et al., 2003). Patients with SZ were found to either commit more errors, or had longer response latency, or both. Significant relationships between HSCT measures and symptoms in individuals with
SZ have also been established (Chan et al., 2010; Waters et al., 2003).
Results of previous study also suggest that patients with SZ probably
showed the most severe impairment on the HSCT task in comparison
with other executive function tasks such as verbal fluency and the
Modified Wisconsin Card Sorting Test (Nathaniel-James et al., 2004).
Average effect sizes for other executive function measures, such as verbal fluency (d = 1.39), the Stroop Color–Word Test (d = 1.22), the
Trail Making Test B (d = 1.07) and the Wisconsin Card Sorting Test
(d = 0.95) have been reported (Heinrichs and Zakzanis, 1998). However, the average effect size for performance on the HSCT, a task in
which patients with SZ have potentially the greatest difficulty, is still
not known.
A small number of studies directly compared the performance on
HSCT in individuals with BD and SZ. Kravariti et al. (2005) compared
the performance on HSCT in 30 BD patients and 30 SZ patients. While
the performance of BD patients in the manic stage was found to resemble those of patients with SZ with thought disorder and/or reality distortion (delusion and hallucination), the performance of BD patients in the
depressive stage was found to be similar to SZ patients with negative
symptoms. In another study, Joshua et al. (2009) compared HSCT performance between 39 patients with SZ and 40 patients with BD (as
well as a healthy control group) on several measures, including the
overall scaled score (which takes into consideration both response
latency and error rate), the Task A scaled score, the Task B scaled
score, response suppression (subtracting Task A response latency from
Task B response latency), the Task B Error scaled score, Type A Error
score in Task B and Type B Error score in Task B. Results of the study suggested that the overall scaled score difference observed between the
two groups was mainly due to an increase in Type B Error in the SZ
group. There was no reliable difference between the two groups on
other HSCT measures. Results from Kravariti et al.'s (2005) and Joshua
et al.'s (2009) study seem to suggest that there are more similarities
than differences between the performance of patients with SZ and BD
on the HSCT. However, more evidence is needed before a firm conclusion can be drawn. Apart from further investigation on this topic using
large samples, a meta-analysis can be used as a good alternative method
to clarify the issue.
Using published studies which compared either/both patients with
BD or SZ with healthy controls, we could compute effect sizes for the
HSCT measures, and then obtain a grand effect size for each of these
measures for both groups. The meta-analytic method also allows us to
directly compare the effect sizes generated by comparing individuals
with BD with healthy controls and those generated by comparing individuals with SZ and healthy controls. Therefore, the present metaanalysis had two aims. The first was to obtain a general profile of performance deficits on HSCT measures in patients with SZ and BD separately.
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
Secondly, we also aimed to compare the effect sizes generated for patients with BD and SZ directly. For any HSCT measures where a grand effect size could not be computed due to small number of available
studies, a qualitative review based on published data was provided
instead.
2. Methods
2.1. Literature search
The flowchart of data extraction for the meta-analysis of each HSCT
measure is shown in Fig. 1. Potential articles used in the meta-analysis
were identified from Elsevier, EBSCOHost (PsychINFO, PsychACTICLE)
and MedLine databases between January 2000 and June 2012. The keywords used were ‘Hayling Sentence Completion Test’ or ‘HSCT’ or ‘executive functioning’ or ‘inhibition’, and ‘schizophrenia’ or ‘schizophrenic’
or ‘bipolar disorder’. Additional articles were obtained from the
155
reference list of the articles identified from the databases. These search
procedures yielded an initial pool of 32 potential articles.
The following inclusion criteria were used to select studies from the
initial article pool for quantitative analysis:
(a) Patients met diagnostic criteria for SZ or BD according to various
versions of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III, DSM-III-R, DSM-IV) or Research Diagnostic
Criteria (RDC);
(b) The HSCT was administered to the patient group(s) (either patients with SZ or BD or both) and healthy controls;
(c) Means and standard deviations (SDs) for at least one of the HSCT
measures were available, or exact t-values/p-values/F-values for
comparing any HSCT measures between patients and controls
were reported (note: one study used median and IQR).
(d) The article was written in English.
The exclusion criteria were:
(a) Studies that included data already reported in a previous study;
(b) Case studies.
32 papers identified by
literature search
17 papers not meeting
inclusion criteria and
were excluded
15 papers retained
2 papers excluded due to
exclusion criteria
Based on these criteria, the final number of articles used in the metaanalysis was 13. Among these, one included a group of BD and a group of
SZ patients (Joshua et al., 2009); one included a group of BD patients in
the manic phase, a group of BD patients in the depression phase, and a
group of BD patients in the remitted phase (Dixon et al., 2004); and
one included a group of patients with SZ with clear negative symptoms
and a group of patients with SZ with clear thought disorders (Kravariti
et al., 2005). For these three articles, each group of patients together
with their corresponding healthy controls was regarded as an independent data set. Altogether, 17 valid data sets were used in the metaanalysis. The profile of studies and data sets included in the metaanalysis are presented in Table 1.
2.2. Meta-analysis procedure
13 papers (17 datasets)
retained
Task A response latency:
6 datasets for BD and 7
datasets for SZ
Task B response latency:
5 datasets for BD and 4
datasets for SZ
Suppression time: 4
datasets for BD and 5
datasets for SZ
Total Error in Task B: 5
datasets for BD and 8
datasets for SZ
Type A error in Task B:
2 datasets for BD and 6
datasets for SZ
Type B error in Task B:
2 datasets for BD and 6
datasets for SZ
Fig. 1. Flowchart for the inclusion of published data for the current meta-analysis.
The meta-analysis was carried out using the Comprehensive
Meta-Analysis software package (Borenstein et al., 2005). Effect
sizes for different HSCT measures were calculated as Cohen's d
which is defined as the difference in group means divided by the
pooled SDs for the two groups. When means and SDs were not
given, the exact t-values, p-values or F-values were used to calculate the effect sizes. The HSCT measures included Total Latency of
Task A, Total Latency of Task B, Suppression Time (subtracting
Total Latency of Task A from Total Latency of Task B), Total Error
of Task B, Type A Error of Task B, Type B Error of Task A, HSCT total
scaled score, Task A scaled score, and Task B scaled score. Shorter latency and fewer errors in Task A indicate better initiation and
shorter latency and fewer errors in Task B indicate better inhibition.
According to Cohen (1992), a small effect size was defined as d =
0.2–0.5; a medium effect size as d = 0.5–0.8; and a large effect
size as d N 0.8.
If the number of available studies for computing the effect size of a
certain HSCT measure was larger than two, a weighted grand mean
was calculated for each set of effect sizes. Random models were used
throughout the meta-analysis. Q-statistics was calculated to evaluate
the homogeneity of the studies. This statistics is used to test whether
the effect sizes of the studies could be assumed to have come from a single population. A significant Q statistics suggests heterogeneity of the
individual study effect sizes which implies a greater variation between
studies. A fail-safe number was calculated to estimate the number of
unpublished studies with nil or minimal effect sizes required to reduce
an overall effect size to some specified negligible value (Rosenberg,
2005).
156
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
Table 1
Profile of studies and data sets included in the meta-analysis (numbers in parentheses are standard deviations).
Studies
Mean age
Numbers
Patient Number
of patients
of controls
group of patients
(male: female) (male: female)
Mean age
of controls
Chan et al., 2010a,b,c,d,e,f
Chan et al., 2012d,e,f
SZ
SZ
29:12
23:11
15:10
16:18
27.9 (8.5)
27.6 (10.8)
Dixon et al., 2004a,b,c,d
Marczewski et al., 2001a,b,d,e,f
Nathaniel-James et al., 1996a,d
BDD
BDM
BDR
BDR
SZ
BD
SZ
SZ-N
SZ-TD
SZ
SZ
6:9
7:8
8:7
20:24
20:10
16:24
26:13
10:5
11:4
9:6
21:4
17:13
17:13
17:13
20:24
30:42
22:22
22:22
17:13
17:13
15 in total
25 in total
Nathaniel-James et al., 2004d
De Almeida Rocca et al., 2008a,b,d
Stoddart et al., 2007a,b,d,e,f
Waters et al., 2003a,b,c,e,f
SZ
BD
BD
SZ
20:6
12:13
6:16
35:7
22:16
10:21
18:22
20:4
Frangou et al., 2005a
Groom et al., 2008a,b,d,e,f
Joshua etal., 2009c,e,f
Kravariti et al., 2005a,c,d
28.1 (6.6)
28.1 (9.9)
Mean IQ
in patients
Mean IQ
in controls
86.5 (17.4)
94.5 (13.6)
Verbal:
Verbal:
33.1 (4.5)
29.7 (7.7)
Performance: Performance:
23.1 (4.9)
18.6 (6.6)
33.9 (8.2)
35.2 (9.8)
n.a.
n.a.
34.3 (11.6) 35.2 (9.8)
n.a.
n.a.
35.7 (9.3)
35.2 (9.8)
n.a.
n.a.
42.7 (11)
43.1 (11.2) 105.8 (13.4) 110.1 (13.9)
19.21 (1.70) 17.19 (2.03) 89.34 (13.59) 107.36 (12.61)
42.1 (11.5) 41.0 (12.0) 107.6 (10.3) 111.8 (6.1)
42.3 (10.7) 41.0 (12.0) 109.2 (9.4)
111.8 (6.1)
34.1 (8.1)
35.2 (9.8)
8.5 (2.0)a
10.0 (2.2)a
33.4 (11.0) 35.2 (9.8)
6.1 (2.1)a
10.0 (2.2)a
29.93 (7.2) 30.4 (6.9)
n.a.
n.a.
Matched at individual
Matched at individual
level
level
36.4 (8.5)
27.9 (9.7)
115.7 (8.3)
114.9 (7.1)
33.1 (8.5)
35.4 (7.0)
101.9 (9.4)
97.2 (9.5)
48.7 (9.7)
36.6 (14.3) 115.7 (6.3)
119.5 (4.4)
36.73 (8.41) 34.67 (8.81) 100.2 (9.3)
103.6 (4.8)
Duration of
illness (yrs)
Years of
education
in patients
Years of
education
in controls
5.4 (5.3)
29–
321.5 days
(first episode)
10.2 (3.0)
11.0 (3.0)
11.0 (2.3)
10.6 (1.6)
6.3 (5.8)
10.7 (10.0)
9.3 (8.1)
15.5 (20.3)
n.a.
19.8 (11.6)
18.7 (9.9)
11.1 (8.5)
10.5 (7.8)
8.1 (5.8)
15.6
13.9 (2.6) 12.8 (2.0)
13.0 (2.7) 12.8 (2.0)
15.6 (2.6) 12.8 (2.0)
13.1 (2..9) 14.2 (2.8)
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
12.0 (0.8) 12.8 (2.0)
13.3 (2.8) 12.8 (2.0)
10.6 (2.0) 10.67 (2.0)
Matched at individual
level
n.a.
n.a.
12.58 (1.7) 11.74 (2.0)
n.a.
n.a.
11.0 (2.0) 11.8 (1.9)
n.a.
9.1 (4.9)
n.a.
13.64
Note: a = Total Latency of Task A; b = Total Latency of Task B; c = Suppression time; d = Total Error of Task B; e = Type A Error in Task B; f = Type B Error in Task B; BD = bipolar
disorder; BDD = bipolar disorder in depression phase; BDM = bipolar disorder in manic phase; BDR = bipolar disorder in remission; SZ = schizophrenia; SZ-N = schizophrenia
patients with predominant symptoms of psychomotor poverty; SZ-TD = schizophrenia patients with predominant symptoms of disorganization.
n.a. = information not available.
a
Age scaled vocabulary score measured with Wechsler Adult Intelligence Scale — Revised.
3. Results
Thirteen studies compared the HSCT scores of SZ or BD patients with
healthy controls. The total number of patients included in these studies
was 292 for SZ and 176 for BD. Mean effect sizes and their 95% CIs and
Q-value are presented in Table 2.
3.1. Total Latency of Task A
Of the 13 studies, 13 sets of data were used to compare Total Latency
of Task A. Among them, six data sets compared patients with BD and
healthy controls and seven data sets compared patients with SZ and
healthy controls. As seen in Table 2, for the BD group results, a significant
effect size of 0.719 (0.231, 1.207) was obtained, Z = 2.888, p b 0.01, indicating that patients with BD performed significantly slower than
healthy controls in Task A. However the Q-statistics indicated significant
heterogeneity for the studies, Q(5) = 22.044, p b 0.001. The fail-safe
number of studies was 48. As illustrated in Fig. 2A, the study by
De Almeida Rocca et al. (2008) showed a low-to-medium effect size of
−0.355 in which healthy controls had higher latency scores in Task A
than BD patients. After excluding this paper from the pool, a large effect
size of 0.896 was obtained (0.592, 1.201), and all the studies in the pool
became homogenous (p N 0.05). The fail-safe N value was 59. De
Almeida Rocca et al. (2008) recruited a group of patients with higher
Table 2
Results of meta-analysis of differences in HSCT scores between healthy controls and patients with bipolar disorder and between healthy controls and patients with schizophrenia.
HSCT measure
Patients
N
No. of patients
No. of HC
da
SE
Total Latency of Task A
BD
6
136
205
0.249
SZ
7
184
221
Total Latency of Task B
BD
5
92
161
Suppression time
SZ
BD
4
4
109
84
136
134
0.719⁎
(0.896⁎)
0.749⁎
(0.881⁎)
0.930⁎
(0.674⁎)
0.840⁎
SZ
5
132
153
BD
5
92
161
SZ
BD
SZ
BD
SZ
8
2
6
2
6
181
62
181
62
181
266
84
214
84
214
Total Error of Task B
Type A Error of Task B
Type B Error of Task B
0.156
(0.397⁎)
0.325
(0.460⁎)
0.866⁎
(1.040⁎)
0.944⁎
0.678⁎
0.639⁎
0.869
0.578⁎
(0.706⁎)
Q value
Fail safeb
0.231, 1.207
22.044⁎⁎
48 (59)
0.195
0.367, 1.132
18.695⁎⁎
74 (78)
0.269
0.403, 1.457
14.778⁎
52 (44)
0.140
0.240
0.566, 1.113
−0.313, 0.653
0.414
8.078⁎
33
0 (2)
0.124
−0.065, 0.549
10.038⁎
4 (10)
0.237
0.402, 1.330
11.628⁎
45 (40)
0.125
0.175
0.106
0.685
0.170
0.698, 1.190
0.336, 1.021
0.431, 0.847
−0.472, 2.211
0.247, 0.912
10.018
5.828⁎
163
–
47
–
40 (44)
95% CI
BD: bipolar disorder; CI: confidence interval; d: Cohen's d; HC: healthy control; N: number of data sets; SE: standard error. SZ: schizophrenia.
⁎ p b 0.05.
⁎⁎ p b 0.01.
a
Numbers in parentheses indicate the effect size after excluding outliers.
b
Numbers in parentheses indicate the Fail-safe Number after excluding outliers.
0.639
13.561⁎
12.178⁎
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
A
157
B
Subgroup
Std diff in means
Study
within study
and 95% CI
BD Dixon et al. 2004 B DD
BD Dixon et al. 2004 BDM
BD Dixon et al. 2004 B DR
BD Frangou et al. 2005
BD de Almeida Rocca et al. 2008
BD Stoddart et al.2007
BD GRAND
SZ Chan et al. 2010
SZ Groom et al. 2008
SZ Kravariti et al. 2005 negative
SZ Kravariti et al. 2005 TD
SZ Marczewski et al. 2001
SZ Nathaniel -James et al. 1996
SZ Waters et al. 2003
SZ GRAND
-4
-2
0
2
Subgroup
within study
BD
BD
BD
BD
BD
SZ
SZ
SZ
SZ
4
Study
Std diff in means
and 95% CI
Dixon et al. 2004 BDD
Dixon et al. 2004 B DM
Dixon et al. 2004 B DR
de Almeida Rocca et al. 2008
Stoddart et al. 2007
BD GRAND
Chan et al. 2010
Groom et al. 2008
Marczewski et al. 2001
Waters et al. 2003
SZ GRAND
-4
-2
0
2
4
Fig. 2. Effect sizes for response latency of Task A (2A) and Task B (2B) of Hayling Sentence Completion Test between patients and healthy controls. BD: bipolar disorder; SZ: schizophrenia;
BDD: bipolar disorder in depression phase; BDM: bipolar disorder in manic phase; BDR: bipolar disorder in remitted phase; TD: thought disorder.
IQ and more years of education than controls (though not statistically
significant p = 0.07). This could explain their unexpected findings. A
positive correlation between IQ and Task A scaled score (r = 0.56) has
been reported elsewhere for BD patients (Joshua et al., 2009).
A significant effect size of 0.749 (0.367–1.132) was obtained from
seven data sets with SZ patients, Z = 3.840, p b 0.01. The fail-safe number was 74, which supported the validity of the results. The studies were
heterogeneous, Q(6) = 18.695, p b 0.01. As presented in Fig. 2A, the
study by Chan et al. (2010) produced a small negative effect size of
−0.077. Among these studies, the study by Chan et al. (2010) was the
only study that used Chinese participants. The HSCT is on the one
hand language based and on the other hand is related to language ability
(Task A specifically). In another study carried out in Hong Kong, the researchers did not report any difference in Task A measures between Chinese first-episode SZ patients and healthy controls (Chan et al., 2012).
Chinese is a language that is very different from alphabetic languages
in terms of orthographic form, orthographic–phonologic transformation, the way by which children learn the written language, and so on.
It is therefore highly possible that there was a cultural effect related to
language which affected the results of this study. After excluding the
Chan et al. (2010) study from the pool, the effect size changed to
0.881 (0.583, 1.178). All the studies in the pool were homogeneous,
Q(5) = 7.722 p N 0.05. The fail-safe N value was 78.
A between-group comparison procedure was carried out to compare
the effect sizes obtained in the two clinical groups and their controls.
The effect sizes obtained when comparing SZ patients with healthy controls were not significantly different from effect sizes obtained when
comparing BD patients with healthy controls, Q(1) = 7.722 p N 0.05,
indicating a similar extent of impairment in the two patient groups.
3.2. Total Latency of Task B
Five data sets compared the Total Latency of Task B between BD patients and healthy controls. As seen in Table 2, a large effect size of 0.930
(0.403, 1.457) was obtained, indicating that patients with BD performed
Task B significantly slower than healthy controls. The fail-safe number
was 52. The Q-statistic suggested heterogeneity, Q(4) = 14.778,
p b 0.05. As shown in Fig. 2B, the study by Stoddart et al. (2007) generated an extremely large effect size (1.842). The control group in this
study was significantly younger than the patients (a mean difference
of 12 years), which might be a potential reason for the unusually large
effect size. After excluding the study by Stoddart et al. (2007) from the
pool, the rest of the studies became homogeneous and produced an
overall effect size of 0.674.
Four data sets compared the Total Latency of Task B between SZ patients and healthy controls and generated a significant effect size of
0.840 (0.566–1.113), Z = 6.014, p b 0.001. The large effect size indicated
that patients with SZ were slower than healthy controls when
performing Task B. The effect sizes of the four studies were homogenous,
Q(3) = 0.414, p N 0.05.
The effect sizes obtained between patients with SZ and healthy controls were similar to those obtained between BD patients and controls,
Q(1) = 0.066, p N 0.05, as indicated by the between-group comparison
procedure. This suggests that the extent of impairment in Task B latency
was similar in the two patient groups.
3.3. Suppression time
Four data sets compared suppression time between patients with BD
and healthy controls, and five data sets compared suppression time between patients with SZ and healthy controls. Although patients with
BD had a longer suppression time, the overall effect size of 0.156
(−0.313, 0.653) was not significant, Z = 0.654, p N 0.05. The studies
were heterogeneous, Q(3) = 8.078, p b 0.05. On inspection of Fig. 3A,
BD patients in the depressive phase in the study by Dixon et al. (2004)
produced an effect size of −0.574. After excluding this study from
the pool, a significant effect size of 0.397 was obtained, Z = 2.517,
p = 0.01. The studies were homogenous (Q(2) = 0.754, p N 0.05), but
the fail-safe number was only 2. These results indicated that patients
with BD took a longer time to suppress a pre-potent response, except patients in the depressive phase who needed less time to do so.
A non-significant effect size of 0.325 was obtained from the five data
sets that compared patients with SZ and healthy controls, Z = 1.627,
p N 0.05. The Q-statistic indicated significant heterogeneity of the studies, Q(4) = 10.038, p b 0.05. As shown in Fig. 3A, the study by Chan
et al. (2010) generated an effect size of −0.398. When this study was
excluded, the other studies became homogeneous, Q(3) = 3.327,
p N 0.05, and a low-to-moderate effect size of 0.460 was obtained,
which was significant, Z = 3.20, p = 0.001. The fail safe N value was
10. The unique feature of the study by Chan et al. (2010) is that the researchers recruited a group of Chinese participants to undertake the
HSCT in Chinese. Overall, the results showed that patients with SZ
took longer time to suppress a strong response, except Chinese patients
who took a little less time to do so.
The effect sizes obtained between patients with SZ and healthy controls were similar to those obtained between BD patients and controls,
Q(1) = 0.202, p N 0.05, as indicated by the between-group comparison
procedure. This suggests that the extent of impairment in the two patient groups was similar.
3.4. Total Error of Task B
Five data sets compared the Total Error of Task B between BD patients and healthy controls. A significant effect size of 0.866 (0.402,
158
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
A
B
Subgroup
within study
BD
BD
BD
BD
SZ
SZ
SZ
SZ
SZ
Study
Subgroup
within study
Std diff in means
and 95% CI
B
B
B
B
B
Dixon et al. 2004 BDR
Dixon et al. 2004 BDD
Dixon et al. 2004 BDM
Joshua et al. 2009
BD GRAND
Chan et al. 2010
Joshua et al. 2009
Kravariti et al. 2005 Negative
Kravariti et al. 2005 TD
Waters et al. 2003
SZ GRAND
SZ
SZ
SZ
SZ
SZ
SZ
SZ
SZ
-2
-1
0
1
2
Study
Std diff in means
and 95% CI
Dixon et al. 2004 BDD
Dixon et al. 2004 BDM
Dixon et al. 2004 BDR
de Almeida Rocca et al. 2008
Stoddart et al. 2007
BD GRAND
Chan et al. 2010
Chan et al.2012
Groom et al. 2008
Kravariti et al. 2005 Negative
Kravariti et al. 2005 TD
Marcewski et al. 2001
Nathaniel-James et al. 1996
Nathaniel-James et al. 2004
SZ GRAND
-4.0 -2.0 0.0
2.0
4.0
Fig. 3. Effect sizes for suppression time (subtracting Task A response latency from Task B response latency) (3A) and the number of Total Error in Task B (3B) of Hayling Sentence Completion Test between patients and healthy controls. BD: bipolar disorder; SZ: schizophrenia; BDD: bipolar disorder in depression phase; BDM: bipolar disorder in manic phase; BDR: bipolar
disorder in remitted phase; TD: thought disorder.
1.330) was obtained, Z = 3.656, p b 0.001. The effect sizes of these
studies were not homogeneous (Q(4) = 11.628, p b 0.01). As shown
in Fig. 3B, the heterogeneity could mainly be attributed to the small effect size in the study by De Almeida Rocca et al. (2008) in which BD patients had more years of education and higher IQ than healthy controls
(though not statistically significant; p = 0.07). After excluding this
study from the pool, the effect sizes of the four remaining studies became homogeneous, Q(3) = 5.162, p N 0.05, generating a significant
effect size of 1.040 (0.626–1.454), Z = 4.25, p b 0.001, with a fail-safe
number of 40.
Eight data sets were included to compare the Total Error of Task B
between patients with SZ and healthy controls. A large effect size of
0.944 (0.698, 1.190) was obtained, Z = 7.531, p b 0.01, indicating that
SZ patients made significantly more errors than healthy controls. All
the studies were homogeneous, Q(6) = 10.018, p N 0.05. The fail-safe
number was 163.
The effect sizes obtained between patients with SZ and controls
were not significantly different from those obtained between BD patients and controls, Q(1) = 0.085, p N 0.05, as indicated by the
between-group comparison procedure. Fig. 3B summarizes the studies
used for computing effect sizes, the effect sizes and confidence intervals.
3.5. Type A Error in Task B
Only two data sets were available for computing the effect size for
Type A Error in Task B between patients with BD and healthy controls,
producing a significant effect size of 0.678 (0.366, 1.021), Z = 3.881,
p b 0.05. The two studies were heterogeneous (Q(1) = 5,828, p b0.05).
As illustrated in Fig. 4A, the study by Stoddart et al. (2007) generated
an unusually large effect size as compared with the other study. The
A
significantly younger control group might be the reason for the heterogeneous result. The fail-safe number could not be computed due to the
small number of studies available.
A medium effect size of 0.639 (0.431, 0.847) was obtained from
the six data sets comparing patients with SZ and healthy controls,
Z = 6.013, p b 0.01. The studies were homogenous, Q(5) = 0.639,
p N 0.05. The fail-safe number was 47. The effect sizes generated by
comparing SZ with healthy controls were similar to those generated
by comparing BD with healthy controls, Q(1) = 0.089, p N 0.05. Thus,
patients with both SZ and BD made significantly more Type A error
during task inhibition to a similar extent.
3.6. Type B Error in Task B
Only two data sets were available for computing the effect size between BD and healthy controls. An effect size of 0.869 was obtained,
but it was not significant, Z = 1.270, p N 0.05. The two studies were
highly heterogeneous, Q(1) = 13.561, p b 0.001. The study by Stoddart
et al. (2007) generated a significantly larger effect size than the study
by Joshua et al. (2009). The fail-safe number could not be calculated because there were only two studies.
Six data sets compared patients with SZ and healthy controls on this
measure. A significant effect size of 0.578 (0.247, 0.912) was obtained,
Z = 3.415, p = 0.001. The results were heterogeneous, Q(5) = 12.178,
p b 0.05. The fail-safe number was 33. The study by Chan et al. (2010)
with chronic Chinese SZ patients reported an effect size of −0.152,
indicating a relatively normal performance in the patient group. After excluding this study from the pool, the overall effect size changed to 0.706,
Z = 5.504, p b 0.001. The four remaining studies were homogenous,
Q(4) = 4.938, p N 0.05, with a fail-safe number of 44. Another study in
B
Subgroup
Study
within study
BD Joshua et al. 2009
BD Stoddart et al. 2007
BP GRAND
SZ Chan et al. 2010
SZ Chan et al. 2012
SZ Groom et al. 2008
SZ Joshua et al. 2009
SZ Marcewski et al. 2001
SZ Waters et al. 2003
SZ GRAND
-2.0
Std diff in means
and 95% CI
Subgroup
within study Study name
BD
BD
-1.0
0.0
1.0
2.0
Joshua et al. 2009
Stoddart et al. 2007
BP GRAND
SZ
Chan et al. 2010
SZ
Chan et al. 2012
SZ
Groom et al. 2008
SZ
Joshua et al. 2009
SZ
Marcewski et al. 2001
SZ
Waters et al. 2003
SZ GRAND
-4.0
Std diff in means
and 95% CI
-2.0
0.0
2.0
4.0
Fig. 4. Effect sizes for the number of Type A Error (4A) and Type B Error (4B) of Task B of Hayling Sentence Completion Test between patients and healthy controls. BD: bipolar disorder; SZ:
schizophrenia.
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
the pool with first episode Chinese SZ patients did report worse performance in patients as compared to healthy controls (Chan et al., 2012).
The potential difference in the two studies might indicate that performance on the HSCT task could be improved with the stability of symptoms, at least in Chinese patients.
We also compared the effect sizes generated by comparing the two
patient groups with healthy controls and found no significant difference, Q(1) = 0.018, indicating that patients with SZ and patients with
BD were comparable in terms of Type B Error in Task B. However
there was significant heterogeneity among the small number of studies
available for meta-analysis. Fig. 4B shows the effect sizes and confidence
intervals generated.
4. Discussion
Consistent with findings from many previous studies, we found in
this study, that patients with SZ and BD were both impaired on the
HSCT. They were slower in task initiation, as indicated by the medium
effect size in Total Latency in Task A (d = 0.881 for SZ, d = 0.896 for
BD). Compared with other measures, very few studies have measured
errors in Task A in patients with SZ (Chan et al., 2010) and in patients
with BD (de Almeida Rocca et al., 2008; Dixon et al., 2004). The limited
number of studies on this issue seems to suggest a relatively high correct
rate in Task A. Nevertheless, Dixon et al. (2004) did find BD patients in
the manic and remitted phases making more errors in Task A. The combined results suggest that both patient groups used a good strategy for
Task A, namely they slowed their response to generate a correct answer.
This strategy, however, seems insufficient for performing Task B when
additional inhibition and planning are needed. They responded more
slowly and committed more errors in Task B, as indicated by the large effect sizes in Total Latency in Task B (d = 0.840 for SZ, d = 0.674 for BD)
and Total Error of Task B (d = 0.944 for SZ, d = 1.040 for BD). Inhibition
appeared to be difficult for them and as a result they made more Type A
errors (d = 0.639 for SZ, d = 0.678 for BD) and Type B errors (d =
0.706 for SZ, d = 0.869 for BD). The findings from the present metaanalysis, together with results from other meta-analyses on executive
functions in BD and/or SZ (Fioravanti et al., 2005; Heydebrand, 2006;
Laws, 1999; Reichenberg, 2010), confirm that executive functions are
impaired in both patient groups.
Some researchers have associated impaired inhibition on the HSCT
with working memory function (Royer et al., 2009b). To what extent
is the observed impaired inhibition in patients contributed to by deficiency in working memory is an interesting topic. This is because deficient working memory capability in both patient groups has been well
documented (Arts et al., 2008; Reichenberg, 2010). One study that
used Chinese SZ participants may be relevant here. Compared to the
same materials in other languages, Chinese sentences had fewer syllables. Therefore, listening to the same sentence in Chinese might be
less taxing on phonological working memory than in an alphabetic language such as English. Chan et al. (2010) found that there was no reliable difference in HSCT performance between a group of Chinese
patients with chronic SZ and healthy controls, even though the patients
performed worse than healthy controls in a Stroop-like task. Based on
the findings of Chan et al. (2010), we speculate that due to the difference in language, Chinese patients with chronic SZ may expend less
working memory in context building and have more working memory
reserve for both initiation and inhibition. Compared with healthy controls, patients with SZ did not utilize strategies which were commonly
used by healthy controls, such as providing answers with items around
the room or all from one particular semantic category (Joshua et al.,
2009). Patients with SZ or BD seem not to have sufficient working memory resource to launch such a strategy for the task. Further studies are
needed to clarify the relationship between inhibition and working
memory. In a recent longitude study on a group of SZ patients in Hong
Kong, researchers found that patients performed worse during the
159
first episode, but in the third year after onset, the performance in patients became similar to those of healthy controls (Chan et al., 2012).
Findings from the present meta-analysis suggest that patients
with SZ and patients with BD are impaired to a similar extent on the
HSCT. It has been suggested that executive dysfunctions might be a potential marker of familial vulnerability to SZ and BD (Zalla et al., 2004).
We also notice that previous meta-analyses comparing the executive
functions of patients with both disorders found that patients with SZ
are moderately more impaired in executive functions than patients
with BD (Daban et al., 2006; Krabbendam et al., 2005). This may be related to the different tasks used in different studies. For example, the
Trail Making Test (Reitan, 1958) and the Stroop Task (Stroop, 1935)
were used to assess executive functions in the meta-analysis by
Krabbendam et al. (2005). Even though both tests are widely used to assess executive functions, it is possible that the two tests target different
domains of executive functions than the HSCT, which mainly targets inhibition function. While the Trail Making Test focuses on task switching
and sequencing (Gray, 2006), the Stroop task examines interference
and control of attention (Cohen et al., 1990). As pointed out by
Frangou et al. (2005a), while the correct answer in the Stroop task relates simply to the colors presented, the correct answer in the HSCT requires self-generation as well as planning. Therefore it is not surprising
that patients with BD are more impaired in the HSCT than in the Stroop
task.
In conclusion, patients with SZ and BD both showed impaired performance on the HSCT. They had problems in both task initiation and
task inhibition. The extent of the impairment is similar in the two
groups of patients on many HSCT measures. The relationship between
working memory and disinhibition in both patient groups needs to be
further studied. One limitation of the present meta-analysis is the comparatively small number of studies included. Another limitation is that
we were unable to compare different subtypes of patients due to the
small number of studies.
Acknowledgments
This study was supported partly by the National Key Technologies
R&D Program (2012BAI36B01), the Knowledge Innovation Project of
the Chinese Academy of Sciences (KSCX2-EW-J-8), the Key Laboratory
of Mental Health, Institute of Psychology, and a grant from the National
Science Fund China Young Investigator Award (81088001) to Raymond
Chan. This study was also supported by National Natural Science
Foundation of China (31200775) and the Scientific Foundation of Institute of Psychology, Chinese Academy of Sciences (Y1CX281005) to Kui
Wang, and by an Australian Academy of Science Visit to China grant
and a Queensland International Fellowship to David Shum. Finally,
it was also supported by a grant from the initiation fund of the CAS/
SAFEA International Partnership Programme for Creative Research
Team (Y2CX131003).
References
Andreasen NC, O'Leary DS, Cizadlo T, Arndt S, Rezai K, Ponto LL, et al. Schizophrenia and
cognitive dysmetria: a positron-emission tomography study of dysfunctional
prefrontal-thalamic-cerebellar circuitry. Proc Natl Acad Sci U S A. 1996;93:9985–90.
Arts B, Jabben N, Krabbendam L, van Os J. Meta-analyses of cognitive functioning in
euthymic bipolar patients and their first-degree relatives. Psychol Med 2008;38:
771–85.
Borenstein M, Hedges L, Higgins J, Rothstein H, Borenstein M, Hedges L, et al. Comprehensive meta analysis, version 2. Englewood, NJ: Biostat; 2005.
Breton F, Plante A, Legauffre C, Morel N, Ades J, Gorwood P, et al. The executive control of
attention differentiates patients with schizophrenia, their first-degree relatives and
healthy controls. Neuropsychologia 2011;49:203–8.
Burgess PW, Veitch E, de Lacy Costello A, Shallice T. The cognitive and neuroanatomical
correlates of multitasking. Neuropsychologia 2000;38:848–63.
Chan RCK. Dysexecutive symptoms among a non-clinical sample: a study with the use of
the Dysexecutive Questionnaire. Br J Psychol 2001;92:551–65.
Chan RCK, Chen EYH. Executive dysfunctions and neurological manifestations in schizophrenia. Hong Kong J Psychiatry 2004;14:2–6.
160
K. Wang et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 46 (2013) 153–160
Chan RCK, Chen EYH, Cheung EFC, Cheung HK. Executive dysfunctions in schizophrenia —
relationships to clinical manifestation. Eur Arch Psychiatry Clin Neurosci 2004;254:
256–62.
Chan RCK, Huang J, Guo L, Cao XY, Hong XH, Gao Z. Executive control in schizophrenia in
task involving semantic inhibition and working memory. Psychiatry Res 2010;179:
259–66.
Chan KK, Xu JQ, Liu KC, Hui CL, Wong GH, Chen EY. Executive function in first-episode
schizophrenia: a three-year prospective study of the Hayling Sentence Completion
Test. Schizophr Res 2012;135:62–7.
Christodoulou T, Lewis M, Ploubidis GB, Frangou S. The relationship of impulsivity to response inhibition and decision-making in remitted patients with bipolar disorder.
Eur Psychiatry 2006;21:270–3.
Cohen J. A power primer. Psychol Bull 1992;112:155–9.
Cohen JD, Dunbar K, McClelland JL. On the control of automatic processes: a parallel distributed processing account of the Stroop effect. Psychol Rev 1990;97:332–61.
Daban C, Martinez-Aran A, Torrent C, Tabares-Seisdedos R, Balanza-Martinez V, SalazarFraile J, et al. Specificity of cognitive deficits in bipolar disorder versus schizophrenia.
A systematic review. Psychother Psychosom 2006;75:72–84.
de Almeida Rocca CC, de Macedo-Soares MB, Gorenstein C, Tamada RS, Isller CK, Dias RS,
et al. Verbal fluency dysfunction in euthymic bipolar patients: a controlled study.
J Affect Disord 2008;107:187–92.
Dixon T, Kravariti E, Frith C, Murray RM, McGuire PK. Effect of symptoms on executive
function in bipolar illness. Psychol Med 2004;34:811–21.
Fioravanti M, Carlone O, Vitale B, Cinti ME, Clare L. A meta-analysis of cognitive deficits in
adults with a diagnosis of schizophrenia. Neuropsychol Rev 2005;15:73–95.
Frangou S. Cognitive function in early onset schizophrenia: a selective review. Front Hum
Neurosci 2010;3:79.
Frangou S, Donaldson S, Hadjulis M, Landau S, Goldstein LH. The Maudsley Bipolar Disorder Project: executive dysfunction in bipolar disorder I and its clinical correlates. Biol
Psychiatry 2005a;58:859–64.
Frangou S, Haldane M, Roddy D, Kumari V. Evidence for deficit in tasks of ventral, but not
dorsal, prefrontal executive function as an endophenotypic marker for bipolar disorder. Biol Psychiatry 2005b;58:838–9.
Gray R. Comprehensive Trail Making Test. J Psychoeduc Assess 2006;24:88–91.
Groom MJ, Jackson GM, Calton TG, Andrews HK, Bates AT, Liddle PF, et al. Cognitive deficits in early-onset schizophrenia spectrum patients and their non-psychotic siblings:
a comparison with ADHD. Schizophr Res 2008;99:85–95.
Heinrichs RW, Zakzanis KK. Neurocognitive deficit in schizophrenia: a quantitative review of the evidence. Neuropsychology 1998;12:426–45.
Heydebrand G. Cognitive deficits in the families of patients with schizophrenia. Curr Opin
Psychiatry 2006;19:277–81.
Joshua N, Gogos A, Rossell S. Executive functioning in schizophrenia: a thorough examination of performance on the Hayling Sentence Completion Test compared to psychiatric and non-psychiatric controls. Schizophr Res 2009;114:84–90.
Krabbendam L, Arts B, van Os J, Aleman A. Cognitive functioning in patients with schizophrenia and bipolar disorder: a quantitative review. Schizophr Res 2005;80:137–49.
Kravariti E, Dixon T, Frith C, Murray R, McGuire P. Association of symptoms and executive
function in schizophrenia and bipolar disorder. Schizophr Res 2005;74:221–31.
Laws KR. A meta-analytic review of Wisconsin Card Sort studies in schizophrenia: general
intellectual deficit in disguise? Cogn Neuropsychiatry 1999;4:1–30. (discussion
31–5).
Marczewski P, Van der Linden M, Larøi F. Further investigation of the Supervisory Attentional System in schizophrenia: planning, inhibition, and rule abstraction. Cogn
Neuropsychiatry 2001;6(3):175–92.
Morey RA, Inan S, Mitchell TV, Perkins DO, Lieberman JA, Belger A. Imaging frontostriatal
function in ultra-high-risk, early, and chronic schizophrenia during executive processing. Arch Gen Psychiatry 2005;62:254–62.
Nathaniel-James DA, Brown R, Ron MA. Memory impairment in schizophrenia: its' relationship to executive function. Schizophr Res 1996;21:85–96.
Nathaniel-James DA, Brown RG, Maier M, Mellers J, Toone B, Ron MA. Cognitive abnormalities in schizophrenia and schizophrenia-like psychosis of epilepsy. J Neuropsychiatry Clin Neurosci 2004;16:472–9.
Quraishi S, Frangou S. Neuropsychology of bipolar disorder: a review. J Affect Disord
2002;72:209–26.
Reichenberg A. The assessment of neuropsychological functioning in schizophrenia.
Dialogues Clin Neurosci 2010;12:383–92.
Reitan RM. Validity of the Trail Making Test as an indicator of organic brain damage.
Percept Mot Skills 1958;8:271–6.
Rosenberg MS. The file-drawer problem revisited: a general weighted method for calculating fail-safe numbers in meta-analysis. Evolution 2005;59:464–8.
Royall DR, Mahurin RK, True JE, Anderson B, Brock IP, Freeburger L, et al. Executive impairment among the functionally dependent — comparisons between schizophrenic
and elderly subjects. Am J Psychiatry 1993;150:1813–9.
Royer A, Schneider FC, Grosselin A, Pellet J, Barral FG, Laurent B, et al. Brain activation
during executive processes in schizophrenia. Psychiatry Res 2009a;173:170–6.
Royer A, Schneider FCG, Grosselin A, Pellet J, Barral FG, Laurent B, et al. Brain activation during executive processes in schizophrenia. Psychiatr Res-Neuroimaging 2009b;173:
170–6.
Shallice T. From neuropsychology to mental structure. Cambridge: Cambridge Univ. Press;
1988.
Stoddart SD, Craddock NJ, Jones LA. Differentiation of executive and attention impairments in affective illness. Psychol Med 2007;37:1613–23.
Stroop JR. Studies of interference in serial verbal reactions. J Exp Psychol 1935;18:643–62.
Waters FA, Badcock JC, Maybery MT, Michie PT. Inhibition in schizophrenia: association
with auditory hallucinations. Schizophr Res 2003;62:275–80.
Zalla T, Joyce C, Szoke A, Schurhoff F, Pillon B, Komano O, et al. Executive dysfunctions as
potential markers of familial vulnerability to bipolar disorder and schizophrenia. Psychiatry Res 2004;121:207–17.