Author's personal copy

Linear and non-linear associations of symptom dimensions and cognitive function in

fi

rst-onset psychosis

Eugenia Kravariti

a

,

⁎

,

1

, Manuela Russo

a

,

1

, Evangelos Vassos

a

, Kevin Morgan

a

,

2

, Paul Fearon

a

,

3

,

Jolanta W. Zanelli

a

, Arsime Demjaha

a

, Julia M. Lappin

a

, Elias Tsakanikos

b

,

4

, Paola Dazzan

a

, Craig Morgan

a

,

Gillian A. Doody

c

, Glynn Harrison

d

, Peter B. Jones

e

, Robin M. Murray

a

, Abraham Reichenberg

a

a

NIHR Biomedical Research Centre for Mental Health, South London and Maudsley NHS Foundation Trust and Institute of Psychiatry, King's College London, De Crespigny Park,

London SE5 8AF, UK

b

ESTIA Centre/Health Service and Population Research, South London and Maudsley NHS Foundation Trust and Institute of Psychiatry, King's College London, 66 Snowfields,

London SE1 3SS, UK

c

Community Health Sciences, Queen's Medical Centre, Institute of Clinical Research, University of Nottingham, Nottingham NG7 2UH, UK

d

Academic Unit of Psychiatry, Cotham House, University of Bristol, Bristol BS6 6JL, UK

e

Department of Psychiatry, University of Cambridge, Box 189, Addenbrooke's Hospital, Cambridge CB2 2QQ, UK

a b s t r a c t

a r t i c l e i n f o

Article history:
Received 20 September 2011
Received in revised form 9 May 2012
Accepted 5 June 2012
Available online 4 July 2012

Keywords:
Population based
First onset psychosis
Affective
Non affective
Symptom dimensions
Cognition

Background:

Associations between symptom dimensions and cognition have been mainly studied in non-

affective psychosis. The present study investigated whether previously reported associations between cognition
and four symptom dimensions (reality distortion, negative symptoms, disorganisation and depression) in non-
affective psychosis generalise to a wider spectrum of psychoses. It also extended the research focus to mania, a
less studied symptom dimension.
Methods:

Linear and non-linear (quadratic, curvilinear or inverted-U-shaped) associations between cognition

and the above five symptom dimensions were examined in a population-based cohort of 166 patients with
fi

rst-onset psychosis using regression analyses.

Results:

Negative symptoms showed statistically significant linear associations with IQ and processing speed,

and a significant curvilinear association with verbal memory/learning. Significant quadratic associations
emerged between mania and processing speed and mania and executive function. The contributions of mania
and negative symptoms to processing speed were independent of each other. The findings did not differ
between affective and non-affective psychoses, and survived correction for multiple testing.
Conclusions:

Mania and negative symptoms are associated with distinct patterns of cerebral dysfunction in first-

onset psychosis. A novel finding is that mania relates to cognitive performance by a complex response function
(inverted-U-shaped relationship). The associations of negative symptoms with cognition include both linear
and quadratic elements, suggesting that this dimension is not a unitary concept. These findings cut across
affective and non-affective psychoses, suggesting that different diagnostic entities within the psychosis
spectrum lie on a neurobiological continuum.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Individuals with the same diagnosis within the psychosis spec-

trum often vary considerably in clinical characteristics (

Jablensky,

2006; Joyce and Roiser, 2007; Stroup, 2007

). At the same time,

different diagnostic categories show overlapping psychopathology,
indistinct clinical boundaries and shared etiological factors (

Squires

and Saederup, 1991; Murray et al., 2004; Kaymaz and Van Os,
2009

). Attempts to reconcile the heterogeneity within, and overlap

across, psychoses have considered dimensional (e.g. symptom) ap-
proaches to classification as a useful adjunct or alternative to categor-
ical (e.g. diagnostic) representations. Exploratory factor analyses in
schizophrenia and, more recently, in the full spectrum of psychoses,
have identified a discrete number of psychopathological dimensions
(e.g. psychomotor poverty, disorganisation, reality distortion, mania,
depression) (

Liddle, 1987; Dikeos et al., 2006; Demjaha et al., 2009

).

These have been reported to provide more meaningful information
than diagnostic categories in relation to clinically and neurobiologically
significant characteristics, including disease course/outcome, likely

Schizophrenia Research 140 (2012) 221–231

⁎

Corresponding author. Tel.: [phone]; fax: [phone].
E-mail address:

[e-mail]

(E. Kravariti).

1

Contributed equally to this manuscript (Joint First Authors).

2

Present address: Department of Psychology, University of Westminster, 309 Regent

Street, London W1B 2UW, UK.

3

Present address: Department of Psychiatry, St. Patricks University Hospital and

Trinity College, University of Dublin, James St., Dublin 8, Ireland.

4

Present address: Department of Psychology, Roehampton University, Holybourne

Avenue, London SW15 4JD, UK.

0920-9964/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:

10.1016/j.schres.2012.06.008

Contents lists available at

SciVerse ScienceDirect

Schizophrenia Research

j o u r n a l h o me p a g e : w ww . e l s e v i e r . c o m / l oc a t e / s c h r e s

Author's personal copy

response to treatment (

Van Os et al., 1999; Allardyce et al., 2007

) and

cognitive performance (

Dixon et al., 2004; Simonsen et al., 2009

).

Despite similar symptom dimensions emerging in factor analytic

studies of patients with affective- and those with non-affective psycho-
ses (

Peralta et al., 1997; Lindenmayer et al., 2008; Smith et al., 2009

),

studies exploring relationships between symptom dimensions and neu-
ropsychological performance have mainly focused on non-affective
psychoses, especially schizophrenia. A recent meta-analysis of this liter-
ature reported modest, statistically significant, and partly dissociable
correlations of negative symptoms and disorganisation with neuropsy-
chological performance, but no significant associations of the positive
and depressive symptom clusters with cognition (

Dominguez et al.,

2009

). Compared to disorganisation, negative symptoms yielded a signif-

icantly stronger correlation with verbal fluency, and significantly less ro-
bust associations with reasoning/problem solving and attention/vigilance.
The two dimensions did not differ in their strength of correlation with IQ,
executive control, speed of processing, verbal working memory, and
verbal/visual learning (

Dominguez et al., 2009

). The latter systematic

review identified only four studies exploring the association of cognitive
performance with the manic/excitement dimension and excluded the
corresponding data as being too limited for an informative synthesis.

The present study addressed the hypothesis that findings from

non-affective psychosis in relation to symptom dimensions and neu-
ropsychological performance (specifically, the partly dissociable,
significant associations of negative symptoms and disorganisation
with cognition, and the non-significant associations of reality distortion
and depression with cognition) (

Dominguez et al., 2009

), replicate in a

population-based cohort of patients with first-onset psychoses includ-
ing both non-affective and affective categories. A further aim was to
extend previous findings by exploring associations between neuro-
cognition and manic symptoms.

Although exploring non-linear (quadratic, curvilinear or inverted-

U-shaped) associations between psychopathology and cognition was
not among the aims of the study, in line with strong statistical evi-
dence of nonlinear processes in brain dysfunction in schizophrenia
(

Breakspear, 2006

), our main analysis suggested potential deviations

from linearity for some associations. It was therefore important to fol-
low this indicative finding with post-hoc analyses, particularly in the
light of evidence that many relationships in behavioural and social
sciences do not follow a straight line. Nonlinear curve fitting is often
required in the analysis of biological, biochemical and pharmacologi-
cal data (

Breakspear, 2006

), but is less commonly applied to cognition

and symptom dimensions. An earlier study of recent-onset schizo-
phrenia reported quadratic associations between negative symptoms
and several neuropsychological measures (

Van der Does et al., 1993

),

further underlining the importance of exploring non-linear patterns
in our data.

2. Method

2.1. The ÆSOP study

The data were derived from the baseline population-based ÆSOP

(Aetiology and Ethnicity in Schizophrenia and Other Psychoses) study,
which identified all cases aged 16–64 years with first-onset psychoses
(ICD-10 codes F10-F29 and F30-F33 in ICD-10) (

World Health

Organization, 1992

) presenting to specialist mental health services in

tightly defined catchment areas in South-east London, Nottingham
and Bristol in September 1997–August 2000. Exclusion criteria were
previous contact with health services for psychosis, organic causes of
psychosis, and transient psychotic symptoms due to acute intoxication.
The study further included a random sample of community controls,
and was approved by local research ethics committees. Participants
gave written informed consent to participate. A detailed overview of
the ÆSOP study has been published elsewhere (

Fearon et al., 2006;

Morgan et al., 2006

).

2.2. The analytic cohort

The analytic cohort comprised 166 ÆSOP cases with consensus ICD-

10 diagnoses of schizophrenia (F20; N=64), schizoaffective disorder
(F25; N=8), bipolar disorder/mania (F30.2/F31.2/F31.5; N=31), de-
pressive psychosis (F32.3/F33.3; N=27) or other psychotic disorders,
including persistent delusional, acute and transient, other nonorganic,
and unspecified nonorganic psychotic disorders (F22/F23/F28/F29;
N

=36). All patients had Item Group Checklist (IGC) ratings on the

Schedules for Clinical Assessment in Neuropsychiatry (SCAN) (

World

Health Organization, 1994

), Wechsler Adult Intelligence Scale-Revised

(WAIS-R) (

Wechsler, 1981

) Full-Scale IQ

≥70, one or more measure-

ments on the ÆSOP neuropsychological battery, and a good command
of English. To satisfy the latter criterion, participants were required to
be native speakers of English or migrants to the UK by age 11 (i.e. have
completed at least their secondary education in the UK).

Due to being selected for having no learning disability and for being

proficient in English (which are standard requirements for neuropsy-
chological testing), as expected, the study sample differed significantly
in IQ (t=4.12, d.f.=190, Pb0.001) and ethnicity (χ

2

=20.97,

Pb

0.001) from the remaining ÆSOP cases (of the 370 patients with

IGC ratings who were excluded from the current study, 288–318 had
available demographic and clinical data, and 26 had available IQ data).
The study sample also scored lower on reality distortion compared to
the remaining ÆSOP cases (t=

−2.13, d.f.=482, P=0.033). The two

groups did not differ significantly in gender (χ

2

=0.001, P=0.980),

education (χ

2

=4.916, P=.086), diagnostic breakdown (χ

2

=8.108,

P

=0.088), age at testing (t=1.030, d.f.=482, P=0.304), age at illness

onset (t=1.102, d.f.=466, P=0.271), duration of untreated psychosis
(t = 0.263, d.f. = 468, P = 0.793), or dimension scores for mania
(t =

−0.327, d.f. =482, P=0.744), negative symptoms (t =

−1.220, d.f.=482, P=0.223), depression (t=−0.977, d.f. =482,

P

= 0.329) and disorganisation (t = 0.987, d.f. = 482, P = 0.324).

2.3. Assessment of socio-demographic and clinical characteristics

Data on age, gender, ethnicity and education were collected by

interviews with the participants using the Medical Research Council
Sociodemographic Schedule (

Mallett, 1997

). Clinical data were

collected using the SCAN (

World Health Organization, 1994

). This

incorporates the Present State Examination (PSE) Version 10, which
was used to elicit symptom-related data at presentation. Ratings on
the SCAN are based on clinical interview, case note review and infor-
mation from informants (e.g. relatives or health professionals). ICD-
10 diagnoses were determined using the SCAN data on the basis of
consensus meetings involving one of the principal investigators and
other members of the research team. The kappa scores for indepen-
dent diagnostic ratings ranged from 1.0 for psychosis as a category
to 0.6–0.8 for individual diagnoses. The participants' demographic,
diagnostic and medication characteristics are presented in

Table 1

.

2.4. Symptom dimensions

Based on a factor analytic study by the ÆSOP Study Group (

Demjaha

et al., 2009

), patients were rated on five symptoms dimensions: Mania

(6 IGC items: ‘heightened subjective functioning’, ‘expansive mood’,
‘

expansive delusions and hallucinations’, ‘rapid subjective tempo’,

‘

over-activity’, ‘socially embarrassing behaviour’), Reality Distortion

(6 IGC items: ‘non-affective auditory hallucinations’, ‘non-specific
auditory hallucinations’, ‘experience of disordered form of thoughts’,
‘

delusions of reference’, ‘bizarre delusions and interpretations’, ‘de-

lusions of persecution’), Negative Symptoms (4 IGC items: ‘nonverbal
communication’, ‘poverty of speech’, ‘flat and incongruous affect’,
‘

motor retardation’), Depressive Symptoms (3 IGC items: ‘special fea-

tures of depressed mood’, ‘depressed mood’, ‘depressive delusions
and hallucinations’) and Disorganisation (2 IGC items: ‘incoherent

222

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

speech’, ‘emotional turmoil’). Only items with loadings of at least 0.4
were used to construct the five dimensions, which accounted for 48% of
the total variance in symptoms (Mania 15%; Reality Distortion 11%;
Negative Symptoms 10%; Depressive Symptoms 7%; Disorganisation
5%) (

Demjaha et al., 2009

). Each dimension was rated by summing up

the scores of the individual IGC items under that dimension. Scores for
individual IGC items ranged from 0 (below threshold) to 1 (moderate)
to 2 (severe) depending on the frequency and severity of symptoms.

2.5. Neuropsychological assessment

Five neuropsychological domains were evaluated: ‘Full-scale IQ’

was derived from the WAIS-R Vocabulary, Comprehension, Digit
Symbol and Block Design subtests; ‘Verbal Memory/Learning’ was
assessed using the Rey Auditory Verbal Learning Test (Trials 1–5, 6
and 7) (

Spreen and Strauss, 1991

); ‘Visual Memory’ was examined

using Visual Reproduction (immediate recall) of the Wechsler Mem-
ory Scale–Revised (

Wechsler, 1987

); ‘Executive Function/Working

Memory’ was evaluated using Trail Making-Part B (

Reitan, 1958

),

Letter-Number Span (

Gold et al., 1997

) and Raven's Coloured Pro-

gressive Matrices-Sets AB and B (

Raven, 1995

); and ‘Processing

Speed’ was measured using Trail Making Test-Part A (

Reitan, 1958

)

and WAIS-R Digit Symbol.

With the exception of Full-Scale IQ (which is a standard score

based on population-based normative data, as described in the
WAIS-R manual), normative standards for the neuropsychological
measurements were created by regressing age, gender, ethnicity
and education on each of the neuropsychological variables in 177
healthy ÆSOP controls (

Zanelli et al., 2010

), and then creating stan-

dard (i.e., Z) scores from the regression-adjusted (residual) scores.
A similar procedure was applied to the patient sample (Z-scores
were created using the Mean±SD of the controls' residual scores).
Where appropriate, Z-scores were averaged across tests to give a sin-
gle score per cognitive domain.

2.6. Data analysis

Regression analyses carried out in the programme STATA v.10.0

for Windows (

StataCorp, 2007

) showed no significant interac-

tions between the effects of ‘type of psychosis’ (‘affective’ vs. ‘non-
affective’) and symptom dimensions on neurocognitive function
(all P values > 0.05), with the exception of an interaction between
depression and type of psychosis in relation to IQ (P b0.05) (see

Results

). As mentioned above, unlike the procedure followed in rela-

tion to the remaining cognitive domains, the covariate effects of gen-
der, age, ethnicity and education were not regressed out of IQ. After

Table 1
Demographic, diagnostic, medication and symptom characteristics of patients with first-onset psychoses (n=166).

Patients with first onset psychoses

Total sample (n=166)

Non-affective
psychoses (n=100)

Affective psychoses
(n=66)

Non-affective vs. affective psychoses

N

%

N

%

N

%

Χ

2

d.f.

P

Gender

5.57

1

0.02

Male

94

56.6

64

64.0

30

45.5

Female

72

43.4

36

36.0

36

54.6

Ethnicity

Fisher's Exact

0.19

Caucasian

105

63.3

59

59.0

38

57.6

Caribbean/African

45

27.1

32

32.0

18

27.3

Asian

7

4.2

1

1.0

5

7.6

Other

9

5.4

8

8.0

5

7.6

Completed education

6.93

2

0.03

School

106

63.9

71

71.7

34

52.3

Further

34

20.5

17

17.2

16

24.6

Higher

26

15.7

11

11.1

15

23.1

ICD-10 Disorder

n/a

n/a

n/a

Schizophrenia

64

38.6

64

64.0

0

0.0

Schizoaffective

8

4.8

0

0.0

8

12.1

Bipolar/mania

31

18.7

0

0.0

31

47.0

Depressive psychosis

27

16.3

0

0.0

27

40.9

Other psychosis

36

21.7

36

36.0

0

0.0

Medication

a

Antipsychotic

52

71.2

33

75.0

19

65.5

0.77

1

0.38

Mood Stabilising

8

11.0

0

0.0

7

24.1

Fisher's Exact

0.001

Antidepressant

23

31.5

14

31.8

8

27.6

0.20

1

0.65

Antiparkinsonian

10

13.7

5

11.4

5

17.2

Fisher's Exact

0.51

None

8

11.0

7

15.9

4

13.8

Fisher's Exact

1.00

Mean

SD

Mean

SD

Mean

SD

t

d.f.

P

Age at testing

29.99

10.72

28.84

10.10

31.83

11.36

1.78

164

0.08

Age at illness onset

29.27

10.50

28.06

9.68

31.08

11.46

1.80

160

0.07

DUP (weeks)

57.81

160.12

61.04

126.93

52.59

200.79

0.33

160

0.74

Antipsychotic dose

a

205.50

159.44

188.82

99.30

234.47

230.44

0.82

21.91

0.42

Dimension score

b

Mania

1.63

2.44

0.73

1.10

3.00

3.17

5.60

75.41

b

0.001

Reality distortion

3.34

2.66

3.61

2.88

2.92

2.25

1.72

159.28

0.09

Negative symptoms

1.34

1.81

1.31

1.78

1.38

1.87

0.24

164

0.81

Depression

1.31

1.75

0.91

1.48

1.91

1.95

3.54

112.91

0.001

Disorganisation

0.66

0.91

0.82

1.01

0.41

0.68

3.14

163.92

0.002

a

Information on medication is reported from a sub-sample of 44 cases with non-affective psychoses and 29 cases with affective psychoses, all from London (73 patients, com-

prising 44% of the total patient sample); Nottingham and Bristol did not record detailed information on medication.

b

Each dimension was rated by summing up the scores of the individual Item Group Checklist (IGC) items under that dimension. Scores for individual IGC items ranged from

0 (below threshold) to 1 (moderate) to 2 (severe) depending on the frequency and severity of symptoms. Total dimension scores ranged from 0 to 11 for mania and reality dis-
tortion, from 0 to 8 for negative symptoms, from 0 to 6 for depression and from 0 to 4 for disorganisation.

223

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

co-varying for education (which differed significantly between affective
and non-affective psychoses), this interaction was no longer statistically
significant (P>0.10) (see

Results

). Therefore, the relationship between

each symptom dimension and each neuropsychological domain was
assessed in the total patient sample using univariate regression analyses.

Scatter-plots suggested potential deviations from linearity for some

associations. Therefore, in a second step, we expanded each regression
model (y=a+b

1

*x) with a quadratic term (y=a+b

1

*x+b

2

*x

2

).

When a statistically significant (Pb0.05) non-linear association was

detected, the Likelihood-Ratio (LR) test was performed to examine if
the non-linear model was statistically significantly better fitting than
the linear model (the non-linear model, having more parameters, will
always fit at least as well as the linear model. Whether it fits significant-
ly better and should thus be preferred is determined by deriving the
probability or P‐value of the observed difference D between the two
models when the null hypothesis is true).

Due to the exploratory nature of our analysis and the non-

independence of the neurocognitive domains, we used the False

Table 2
Cognitive scores

a

in patients with first-onset psychoses (n=166).

Patients with first onset psychoses

Total sample
(n=166)

Non-affective
psychoses (n=100)

Affective psychoses
(n=66)

Non-affective vs. affective
psychoses

Mean

SD

Mean

SD

Mean

SD

t

d.f.

P

Full-Scale IQ

b

91.55

16.33

89.86

15.74

94.12

16.98

1.65

164

0.10

Verbal memory/learning

c

−0.64

1.06

−0.67

1.07

−0.60

1.05

0.37

149

0.71

RAVLT Trials 1–5

d

−0.80

1.19

−0.81

1.15

−0.78

1.25

0.14

149

0.89

RAVLT Trial 6

d

−0.55

1.14

−0.58

1.19

−0.49

1.05

0.47

149

0.64

RAVLT Trial 7

d

−0.57

1.08

−0.60

1.09

−0.53

1.07

0.37

145

0.71

Visual memory

e

−0.43

1.01

−0.36

0.96

−0.56

1.08

−1.20

145

0.23

Executive function/working memory

−0.67

0.98

−0.66

0.97

−0.69

1.00

−0.13

156

0.89

Trail Making-Part B

f

−0.55

1.38

−0.45

1.46

−0.70

1.23

−1.09

147

0.28

Letter-Number Span

d

−0.62

1.07

−0.66

1.08

−0.55

1.07

0.61

150

0.54

Raven's CPM-Set AB

d

−0.65

1.48

−0.57

1.35

−0.80

1.67

−0.94

151

0.35

Raven's CPM-Set B

d

−0.79

1.50

−0.83

1.54

−0.73

1.44

0.39

151

0.70

Processing Speed

−0.70

0.98

−0.73

1.01

−0.64

0.93

0.53

158

0.60

Trail Making-Part A

f

−0.66

1.26

−0.69

1.32

−0.62

1.17

0.33

152

0.74

WAIS-R Digit Symbol

d

−0.74

0.94

−0.77

0.96

−0.69

0.91

0.51

155

0.61

Abbreviations: CPM: Coloured Progressive Matrices; RAVLT, Rey Auditory Verbal Learning Test; WAIS-R, Wechsler Adult Intelligence Scale-Revised.

a

With the exception of Full-Scale IQ, all scores are age-, gender-, ethnicity- and education-regressed (residual) raw scores, which were Z-transformed using the mean (SD) scores

of 177 ÆSOP controls (

Zanelli et al., 2010

). Full-Scale IQ is a standard score based on normative data, as described in the Wechsler Adult Intelligence Scale-Revised (WAIS-R) manual

(

Wechsler, 1981

). The mean scores on the cognitive domains of Verbal Memory/Learning, Executive Function/Working Memory and Processing Speed represent averages across the

individual tests that are subsumed under each of the respective cognitive domains.

b

Based on a short form of the WAIS-R (

Wechsler, 1981

) including Vocabulary, Comprehension, Block Design and Digit Symbol. The sums of scaled scores for the Verbal and Per-

formance subtests were prorated by multiplying the sum of the Vocabulary and Comprehension scaled scores by 6/2 and the sum of the Block Design and Digit Symbol scaled scores
by 5/2, respectively. The two prorated sums were summed up before obtaining Full-Scale IQ using the tables in the WAIS-R manual (

Wechsler, 1981

).

c

Based on the number of items recalled correctly in Trials 1–5 (assessing immediate free recall and learning), Trial 6 (assessing short-delay free recall) and Trial 7 (assessing

long-delay free recall) of the Rey Auditory Verbal Learning Test.

d

Total number of correct responses was assessed.

e

Based on the total accuracy score on the immediate recall trial of Visual Reproduction of the Wechsler Memory Scale-Revised (

Wechsler, 1987

), which involves drawing three

geometric designs from memory.

f

Time (seconds) taken to complete the task was assessed.

**/***

Patients with non-affective psychoses differed from those with affective psychoses at the 0.01/0.001 level of

statistical significance

0

0.5

1

1.5

2

2.5

3

3.5

4

Mania

***

Reality Distortion Negative Symptoms

Depression

***

Disorganization

**

Symptom severity

Non-Affective Psychoses

Affective Psychoses

Fig. 1. Symptom dimension scores in patients with non-affective (n=100) and affective (n=66) psychoses.

224

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

Discovery Rate (FDR) method to control for multiple comparisons
(

Benjamini and Hochberg, 1995

). The FDR reflects the proportion of

expected false positives in a set of significant results. The FDR-adjusted
P‐

values are called q-values. The FDR procedure was carried out separate-

ly for the linear- and quadratic-regressionP‐values (5 neurocognitive do-
mains by 5 symptom factors gave rise to 25 P‐values for the linear
regression models and 25 P‐values for the quadratic regression models)
using the bootstrap method and the QVALUE software (

Storey and

Tibshirani, 2003

. Results with P‐valuesb0.05 and q-valuesb0.1 were

retained as significant. A detailed description of the FDR method can be
found in

Curran-Everett (2000)

,

Ling et al. (2009)

and

Strimmer (2008)

.

3. Results

The participants' demographic, diagnostic, medication, symptom

and cognitive characteristics, as well as the results of the statistical com-
parisons between the affective and non-affective categories are pres-
ented in

Tables 1–2

.

Figs. 1–2

display the mean symptom dimension

and cognitive subtest scores in the affective and non-affective patient
groups. The effects of the symptom dimension by group (affective vs.
non-affective) interactions on the cognitive domains are presented in

Table 3

. The results of the linear and non-linear regression models for

the associations between symptom dimensions and cognitive domains
in the full patient sample are presented in

Table 4

and in

Figs. 3–4

.

The affective and non-affective patient groups differed significantly

in gender (Pb0.05), level of completed education (Pb0.05), mood
stabilisers (P=0.001), mania (Pb0.001), depression (P=0.001) and
disorganisation (Pb0.01) scores (

Table 1

,

Fig. 1

), but in none of the cog-

nitive scores (

Table 2

,

Fig. 2

). There was an isolated statistically signifi-

cant group by depression interaction in relation to Full-Scale IQ
(Pb0.05), which disappeared after co-varying for education (

Table 3

).

Statistically significant, both linear and non-linear, associations were

detected in relation to mania and executive function/working memory,
negative symptoms and Full-Scale IQ, negative symptoms and verbal
memory/learning, and negative symptoms and processing speed
(

Table 4

). In addition, a statistically significant non-linear association

was detected in relation to mania and processing speed (

Table 4

). The

q-values indicated low probability (b10%) that these findings were
false positives, with the exception of the linear association between
mania and executive function/working memory (q-value = 0.100).

The Likelihood Ratio test showed that the quadratic model provided

a statistically significantly better fit than the linear model in relation to
mania and executive function/working memory, mania and processing
speed, and negative symptoms and verbal memory/learning (

Table 4

,

Figs. 3–4

).

As both negative symptoms and mania were associated with pro-

cessing speed, we further examined whether these associations were
independent of each other. A multivariate regression model including
both negative symptoms (as a linear term) and manic symptoms (as a
non-linear term, co-varying for the linear term) showed that both di-
mensions were independently associated with processing speed
(negative symptoms: t=

−2.71, P=0.008; manic symptoms: t=

−2.85, P=0.005). Together, the two dimensions explained 10.6% of

the variance in processing speed. In addition, the quadratic associa-
tions between mania and each of the processing speed and executive
function domains remained significant (P=0.002–0.050) after co-
varying for both reality distortion and disorganisation (the three clus-
ters frequently co-exist in the same patient).

In line with published data on the epidemiology of schizophre-

nia (

Howard et al., 2000

), 25 (15.4%) patients of our epidemiologi-

cally ascertained sample had late onset psychoses (>40 years).
After excluding these cases from the analysis, mania and negative
symptoms (but no other symptom dimension) showed statistically
significant associations with the same cognitive domains as in our
main analysis. However, the quadratic model was significantly bet-
ter fitting only in relation to mania and processing speed
(P = 0.007) (and showed a non-significant trend towards a better
fi

t in relation to mania and executive function/working memory;

P

= 0.057).

4. Discussion

Our study examined linear and non-linear associations between five

symptom dimensions and five cognitive domains in an epidemiologically

-1

-0.8

-0.6

-0.4

-0.2

0

RA

VL

T

Tr

ia

ls 1-5

R

AV

L

T

Tr

ial

6

RA

VL

T T

ria

l 7

Vis

ual

R

ep

ro

du

ct

io

n

Trail Ma

kin

g-Pa

rt

B

Le

tte

r-

Nu

m

ber

S

pa

n

Ra

ve

n's CPM-Se

t A

B

Ra

ve

n's CPM-Se

t B

Tra

il M

akin

g-P

ar

t A

WAI

S-

R Dig

it S

ym

bo

l

Verbal Memory/Learning

Visual

Memory

Executive Function/Working Memory Processing Speed

Z scores

Non-Affective Psychoses

Affective Psychoses

Fig. 2. Z scores in 10 neurocognitive tests in patients with non-affective (n=100) and affective (n=66) psychoses.

225

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

Table

3

Re

su

lt

s

of

th

e

re

gr

es

si

on

m

od

els

ex

am

in

ing

st

at

is

ti

ca

li

nt

er

ac

ti

on

s

be

tw

ee

n

ea

ch

sy

m

pt

om

di

m

en

si

on

an

d

ty

pe

of

ps

yc

ho

si

s

(n

on

-a

ff

ec

tiv

e

vs

.a

ff

ec

ti

ve

)

in

re

la

ti

on

to

5

co

gni

tiv

e

do

m

ai

ns

in

pa

tie

nt

s

w

it

h

fir

st

-o

ns

et

ps

yc

ho

se

s

(n

=

16

6)

.

Fu

ll-Sc

ale

IQ

Ver

bal

M

em

or

y/

Le

ar

n

in

g

Vi

su

al

M

em

or

y

Ex

ec

u

ti

ve

Fu

n

ct

io

n

/W

or

k

in

g M

em

or

y

P

ro

ce

ss

in

g S

pe

ed

F

D

.F.

P

F

D

.F.

P

F

D

.F.

P

F

D

.F.

P

F

D

.F.

P

M

an

ia

Linear

1.19

1,162

0.277

0.02

1,147

0.893

2.34

1,143

0.128

0.05

1,154

0.818

0.16

1,156

0.694

Non-Linear

1.09

1,161

0.298

0.52

1,146

0.473

3.26

1,142

0.074

1.32

1,153

0.252

0.37

1,155

0.546

Reality

Distortion

Linear

0.72

1,162

0.396

0.22

1,147

0.642

0.71

1,143

0.400

0.17

1,154

0.682

0.34

1,156

0.559

Non-Linear

0.22

1,162

0.636

0.25

1,147

0.620

0.18

1,143

0.669

0.12

1,154

0.725

0.25

1,156

0.620

N

eg

at

iv

e S

ym

pt

om

s

Linear

0.04

1,162

0.851

0.74

1,147

0.391

2.39

1,143

0.124

0.84

1,154

0.361

2.73

1,156

0.101

Non-Linear

0.00

1,161

0.955

0.29

1,146

0.591

0.68

1,142

0.412

0.35

1,153

0.558

1.46

1,155

0.228

D

ep

re

ss

io

n

Linear

6.58

1,162

0.011

0.21

1,147

0.647

3.03

1,143

0.084

3.65

1,154

0.058

2.81

1,156

0.096

Non-Linear

6.06

1,161

0.015

0.03

1,146

0.874

3.50

1,142

0.064

2.70

1,153

0.102

1.12

1,155

0.292

A

ft

er

c

o-

va

ry

in

g f

or

e

du

ca

ti

on

†

Linear

0.66

1,158

0.418

Non-Linear

0.86

1,157

0.356

Disor

gan

isation

Linear

2.19

1,162

0.140

1.72

1,147

0.192

1.53

1,143

0.218

2.17

1,154

0.142

0.00

1,156

0.975

Non-Linear

0.75

1,161

0.389

2.48

1,146

0.118

0.73

1,142

0.394

1.66

1,153

0.200

0.15

1,155

0.695

† U

nlik

e the pr

oce

dur

e follo

w

ed in r

elation t

o the r

emaining cogniti

ve domains, the co

variat

e ef

fects of

g

ender

, ag

e, e

thnicity

and e

ducation w

er

e no

t r

egr

esse

d out of

F

ull-Scale IQ, r

aising the possibility that the signif

icant Gr

oup b

y

Depr

ession int

er

action in r

elation t

o F

ull-Scale IQ w

as due t

o demogr

aphic dif

fer

ences b

etw

een the af

fecti

ve and non-af

fecti

ve

cat

eg

ories of

psy

chosis.

St

atis

ticall

y signif

icant

P v

alue (<0.05).

226

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

Table

4

Linear

and

non-linear

associations

between

5

symptom

dimensions

and

5

cognitive

domains

in

patients

with

first-onset

psychoses

(n

=

166).

Fu

ll-Sc

ale

IQ

Ver

bal

M

em

or

y/

Le

ar

n

in

g

V

is

u

al

M

em

or

y

Ex

ec

u

ti

ve

Fu

n

ct

io

n

/W

or

k

in

g M

em

or

y

P

ro

ce

ss

in

g S

pe

ed

R

2

P

Q

†

R

2

P

Q

†

R

2

P

Q

†

R

2

P

Q

†

R

2

P

Q

†

M

an

ia

Linear

0.009

0.215

0.003

0.48

6

0.003

0.512

0.034

0.020

0.100

0.015

0.124

Non

-Linear

0.009

0.464

0.004

0.720

0.014

0.362

0.058

0.010

0.071

0.064

0.006

0.058

Linear Vs. No

n-Lin

ea

r ‡

LR

χ

2

(1)

=3.90

, P=0.048

*

LR

χ

2

(1)

=8.21, P

=0.004

**

Reality

Distortion

Linear

0.000

0.851

0.007

0.312

0.001

0.717

0.000

0.939

0.000

0.887

Non

-Linear

0.002

0.851

0.008

0.548

0.003

0.791

0.003

0.773

0.001

0.901

Lin

ea

r

V

s.

N

on

-Li

nea

r

‡

N

eg

at

iv

e S

ym

pt

om

s

Linear

0.048

0.005

0.040

0.050

0.006

0.040

0.000

0.932

0.009

0.229

0.052

0.004

0.040

N

on

-L

in

ea

r

0.

04

9

0.017

0.073

0.074

0.004

0.058

0.005

0.715

0.010

0.457

0.052

0.015

0.073

Li

ne

ar

V

s.

N

on

-L

in

ea

r ‡

LR

χ

2

(1)=0.17, P=0.677

LR

χ

2

(1)=

3.87

, P=0.049

*

LR

χ

2

(1

)=

0.

01

, P

=0

.9

28

D

ep

re

ss

io

n

Linear

0.004

0.426

0.000

0.865

0.006

0.373

0.004

0.420

0.000

0.868

Non-Linear

0.004

0.722

0.000

0.985

0.012

0.424

0.011

0.417

0

.003

0.788

Li

ne

ar

V

s.

N

on

-L

in

ea

r ‡

Disor

gan

isation

Linear

0.010

0.199

0.022

0.068

0.003

0.501

0.000

0.845

0.010

0.213

Non-Linear

0.011

0.398

0.022

0.186

0.011

0.439

0.000

0.967

0.014

0.343

Li

ne

ar

V

s.

N

on

-L

in

ea

r ‡

† Q v

alues ar

e r

eport

ed onl

y for signif

icant

P v

alues (<0.05).

‡ When a s

tatis

ticall

y signif

icant (

P<0.05) non-linear association w

as de

tect

ed, the Lik

elihood-R

atio (LR) t

es

t w

as performe

d t

o e

xamine if

the non-linear model w

as

s

tatis

ticall

y signif

icantl

y b

ett

er f

itting than the linear model (the non-

linear model, ha

ving mor

e par

ame

ters, will alw

ay

s f

it at leas

t as w

ell as the linear model. Whe

ther it f

it

s signif

icantl

y b

ett

er and should thus b

e pr

ef

err

ed is de

termine

d b

y deri

ving the pr

obability or

P-v

alue of

the observ

ed dif

fer

ence D

be

tw

een the tw

o models when the null h

ypo

thesis is true).

The v

alue is s

tatis

ticall

y signif

icant (

P v

alue<0.05) or indicat

es lo

w pr

obability of

a f

alse positi

ve f

inding (Q v

alue<0.

1).

*/** The

P v

alue indicat

es that the non-linear model is s

tatis

ticall

y signif

icantl

y b

ett

er f

itting than the linear model at the 0.05/0.0

1 le

vel of

s

tatis

tical signif

icance.

227

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

ascertained sample of 166 patients with first-onset psychoses, which
included both affective and non-affective diagnoses. In line with our
hypotheses, negative symptoms showed the strongest and most consis-
tent association with cognition, significantly predicting performance in
three of five cognitive domains. These findings are consistent with
those reported in a recent meta-analysis by

Dominguez et al. (2009)

. In

both investigations, negative symptoms predicted deficits in general in-
telligence, verbal memory and processing speed, but not in executive
control or working memory. Further confirming our hypotheses, reality

distortion and depressive symptoms showed weak associations with cog-
nition, which were consistently non-significant, corroborating the broad
fi

ndings by

Dominguez et al. (2009)

. Our study suggests that earlier find-

ings on symptom dimensions and cognitive function in non-affective psy-
chosis have wider applicability to the spectrum of psychoses. Contrary
to our hypothesis and the findings by

Dominguez et al. (2009)

, dis-

organisation failed to elicit significant associations with cognition.

4.1. Negative symptoms and cognition

Although the associations of negative symptoms with intelligence

and processing speed were linear, a significant curvilinear association
was detected in relation to verbal memory/learning. This finding is in
line with an earlier report of quadratic associations between negative
symptoms and several cognitive measures in recent-onset schizo-
phrenia (

Van der Does et al., 1993

). Although the mechanism under-

lying such patterns is not known, the authors speculated that mild
negative symptoms may reflect withdrawal, while a high negative
symptom score is more likely to be indicative of brain pathology
(

Van der Does et al., 1993

). The authors emphasised the distinction

between primary negative symptoms and secondary negative symp-
toms, the latter resulting from depression, neuroleptic medication
or the absence of social stimulation (

Van der Does et al., 1993

). The

replication of non-linear associations between negative symptoms
and cognition in the present study reinforces the view that this symp-
tom dimension is not a unitary concept (

Van der Does et al., 1993

).

Fig. 3. Linear and non-linear associations of Negative Symptoms with‘Full-Scale IQ’, ‘Verbal
Memory/Learning’ and ‘Processing Speed’ in patients with first-onset psychoses (n=166).

Fig. 4. Non-linear associations of Mania with ‘Processing Speed’ and ‘Executive Function/
Working Memory’ in patients with first-onset psychoses (n=166).

228

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

4.2. Mania and cognition

Mania has been little investigated in cognitive/correlational studies

of psychosis to date, and was not examined by

Dominguez et al.

(2009)

. However, the ÆSOP Study Group recently demonstrated that,

of five symptom dimensions, mania showed the highest number of
associations with clinical characteristics and risk indicators (

Demjaha

et al., 2009

). In line with this evidence, mania emerged as the second

most informative dimension in the present analysis, explaining variation
in two of five cognitive domains, i.e. executive function and processing
speed. These associations were independent of those observed in rela-
tion to negative symptoms.

Our study provides novel evidence that mania relates to

neurocognitive performance by a complex response function. The
associations of mania with executive function and processing speed
were inverted-U-shaped, implying that modest levels of mania are
related to better cognitive function compared to minimal or high
levels. A possible explanation is that below a critical threshold, the
excitability that characterises mania boosts the level of motivation
(facilitating engagement with cognitive tasks) and enhances the
responsiveness to cognitive stimuli. This explanation is in line with
the productivity and potential advantages associated with low levels
of mania, countered by the distractibility and impaired decision
making seen with high levels. In addition, the reported associations
between mania and higher pre-morbid IQ, acute mode of onset,
fewer neurological soft signs and shorter duration of untreated
psychosis (

Cannon et al., 2002; Demjaha et al., 2009; Koenen et al.,

2009; MacCabe et al., 2010

) could suggest that individuals with

manic symptomatology are less compromised neurodevelopmentally,
and have better cognitive functioning at baseline (i.e. more cognitive
reserve). It would therefore require higher levels of mania for a
disruptive effect on cognitive performance to become apparent.

4.3. Dimensional views of psychosis

Our statistical analysis showed no evidence of differential associa-

tions between symptom factors and cognitive function in affective
and non-affective psychoses. This finding is in line with earlier
reports (

Kravariti et al., 2005; Simonsen et al., 2009; Smith et al.,

2009

). For example, in a recent study of 72 individuals with schizo-

phrenia and 25 patients with schizoaffective or bipolar disorders,
the two patient groups showed similar dimensions of cognitive func-
tion, similar dimensions of psychopathology, and similar relation-
ships between cognition and symptomathology (

Smith et al., 2009

).

These findings suggest that dimensional or hybrid models of psycho-
sis could prove more useful than categorical models in explaining
neurocognitive performance.

4.4. Methodological strengths

Our study is the first investigation of associations between five

symptom dimensions and five cognitive domains in an epidemiologi-
cally ascertained cohort of patients with a first episode of affective or
non-affective psychosis. It included a broader range of symptom dimen-
sions and diagnostic categories compared to earlier studies. It also ex-
plored the comparability of associations in affective and non-affective
psychoses, and, importantly, it examined whether a non-linear model
offered a more informative account of the explored associations than
a linear pattern. Our preliminary findings suggest that this strategy
may prove fruitful, and that a uniform focus on linear associations
could conceal important relationships between symptom dimensions
and cognition. These novel and robust methodological features enabled
us to address the aims of the study drawing on a uniquely informative
dataset and analysis.

4.5. Methodological limitations

Disorganisation was the least salient factor in the ÆSOP factor an-

alytic study (

Demjaha et al., 2009

), accounting for less variance in

total symptoms (5%) than any other dimension (7%–15%). As the
prominence of disorganisation was critically dependent on the
SCAN and the number of IGC items entered in the analysis, it is possi-
ble that a different clinical schedule might have given rise to a more
salient factor and to significant associations with cognition. In addi-
tion, due to the different number of items (with satisfactory factor
loadings) included in each of the five dimensions, the latter acquired
different score ranges. This caveat, albeit unavoidable, may have
influenced the correlational analysis.

Only patients who met the strict inclusion criteria (e.g. IQ, language,

age of immigration) and could cope effectively with the requirements of
the neuropsychological assessment (those in sub-acute phases) provid-
ed cognitive data to the baseline ÆSOP study. Therefore, only 166 ÆSOP
patients of those with IGC ratings (n=536) were included in the pre-
sent analysis. Although first-episode samples offer many research ad-
vantages (e.g. they are un-confounded by cumulative medication
effects), they also pose research challenges compared to chronic sam-
ples (e.g. their diagnoses may be less reliable and subject to change,
although ‘psychosis’ is generally reliable). Follow-up assessments of
the baseline ÆSOP sample are currently under way and are hoped to
provide interesting comparative data for future analyses.

Due to a lack of non-psychotic affective groups, we could not

examine whether our interesting findings pertaining to mania gener-
alise to mania without psychosis. The cognitive tasks used in the pre-
sent study were not matched for difficulty (for example, letter-
number span makes heavier demands on working memory than
Coloured Progressive Matrices). They further tap complex mental
processes and are likely to load on more than one cognitive domains.
This limitation is inherent in neuropsychological research, and was
addressed by grouping cognitive tasks according to their selective or
prominent, rather than exclusive, properties. The study used an
older edition of the Wechsler intelligence series (WAIS-R), which
may have slightly over-estimated IQ (but is unlikely to have affected
the correlational analysis). The participants were not medication
naïve. Medication may influence symptomatology through treatment
response or side effects, and can impact on the dimensional structure.
Information on medication was only available on 44% of the present
patient sample, and did not include length of time on medication.
This is a limitation of the study, as medication has known effects on
symptoms and cognitive function. It is further important to acknowl-
edge the possibility of multiple disease processes in psychoses, some
driving impairments in routine cognitive and information processing
and some expressing themselves in psychopathology. Finally, correla-
tions between different symptom dimensions and neuropsychologi-
cal performance may be differentially confounded. For example,
acute mania may interfere with neuropsychological test performance,
while negative symptoms may have definitional overlap with some
cognitive aspects.

4.6. Implications and conclusions

In high agreement with earlier reports (

Cameron et al., 2002;

Bozikas et al., 2004; Heydebrand et al., 2004; Lucas et al., 2004

) and

the meta-analysis by

Dominguez et al. (2009)

, the most informative

symptom dimensions (in this study: negative symptoms and mania)
explained a relatively small proportion of variance in neuropsycho-
logical performance (5%–11%). Understanding this replicable finding
is a challenge. The issue skirts the notion of cognitive endo-
phenotypes, i.e. the suggestion that some cognitive deficits tap vul-
nerability to neuropsychiatric disorders, and, as such, are largely
dissociated from symptom states (

Gottesman and Gould, 2003;

Balanzá-Martínez et al., 2008; Burdick et al., 2009; Glahn et al.,

229

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

2010

). Notwithstanding this moderating element, our findings indi-

cate that negative symptoms and mania are associated with different
patterns of cerebral dysfunction, as reflected in discrete patterns of
neuropsychological deficits.

In summary, mania relates to cognitive performance by a complex

response function (inverted-U-shaped relationship) in epidemiologi-
cally ascertained patients with first-onset psychoses. The associations
of negative symptoms with cognition include both linear and quadratic
elements, suggesting that this dimension is also not a unitary concept.
The two dimensions are likely to reflect distinct pathophysiological
processes, which appear to cut across affective and non-affective disor-
ders. Such differences can help understand the heterogeneity of psycho-
ses drawing on dimensional rather than categorical (i.e. diagnostic)
distinctions. Our findings imply that current diagnostic systems may
offer enhanced characterisation of mental disorders by incorporating
dimensional specifications, which can critically inform strategies for
psychiatric rehabilitation.

Role of funding source

This work was supported by the Stanley Medical Research Institute, Bethesda, Md,

which provided financial support for the conduct of study, collection, management and
analysis of data. Evangelos Vassos was supported by a NARSAD Young Investigators
Award.

Contributors

Eugenia Kravariti and Manuela Russo managed the literature search, contributed

to the design and execution of the statistical analysis and wrote the first draft of the
article (Joint First Authors). Evangelos Vassos and Abraham Reichenberg contributed
to the design and execution of the statistical analysis and edited the final manuscript.
All authors contributed to the conceptualization and/or implementation of the study
and edited and approved the final manuscript.

Conflict of interest

Abraham Reichenberg has received speaker's honoraria from AstraZeneca

(Greece). Paola Dazzan has received speaker's honoraria and travel support from
AstraZeneca, Janssen Pharmaceutica, and Sanofi. Peter B. Jones has served as a consul-
tant to Bristol-Myers Squibb, Eli Lilly, and Otsuka. Robin M. Murray has received
speaker's honoraria from AstraZeneca, Janssen Pharmaceutica, Eli Lilly, Bristol-Myers
Squibb, and Novartis Pharmaceuticals. The remaining authors report no financial
relationships with commercial interests.

Acknowledgements

We would like to thank the staff in the mental health services who helped in the

case ascertainment and the research subjects. We gratefully acknowledge advice
from the late R. E. Kendell, FRCPsych, regarding the design of the study. We wish to
acknowledge the contributions of the entire ÆSOP study team, listed online at

http://

www.psychiatry.cam.ac.uk/aesop

.

Appendix A. Supplementary data.

Supplementary data to this article can be found online at

http://

dx.doi.org/10.1016/j.schres.2012.06.008

.

References

Allardyce, J., Suppes, T., Van Os, J., 2007. Dimensions and the psychosis phenotype. Int.

J. Methods Psychiatr. Res. 16 (S1), S34–S40.

Balanzá-Martínez, V., Rubio, C., Selva-Vera, G., Martinez-Aran, A., Sánchez-Moreno, J.,

Salazar-Fraile, J., Vieta, E., Tabarés-Seisdedos, R., 2008. Neurocognitive endo-
phenotypes (Endophenocognitypes) from studies of relatives of bipolar disorder sub-
jects: A systematic review. Special section: The European Workshop in Imagery and
Cognition: Neurocognition and Visual Imagery. Neurosci. Biobehav. Rev. 32 (8),
1426–1438.

Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and

powerful approach to multiple testing. J. R. Stat. Soc. B 57 (1), 289–300.

Bozikas, V.P., Kosmidis, M.H., Kioperlidou, K., Karavatos, A., 2004. Relationship between

psychopathology and cognitive functioning in schizophrenia. Compr. Psychiatry 45
(5), 392–400.

Breakspear, M., 2006. The nonlinear theory of schizophrenia. Aust. N. Z. J. Psychiatry 40

(1), 20–35.

Burdick, K.E., Gunawardane, N., Woodberry, K., Malhotra, A.K., 2009. The role of general

intelligence as an intermediate phenotype for neuropsychiatric disorders. Cogn.
Neuropsychiatry 14 (4–5), 299–311.

Cameron, A.M., Oram, J., Geffen, G.M., Kavanagh, D.J., McGrath, J.J., Geffen, L.B., 2002.

Working memory correlates of three symptom clusters in schizophrenia. Psychia-
try Res. 110 (1), 49–61.

Cannon, M., Caspi, A., Moffitt, T.E., Harrington, H., Taylor, A., Murray, R.M., Poulton, R.,

2002. Evidence for early-childhood, pan-developmental impairment specific to
schizophreniform disorder: results from a longitudinal birth cohort. Arch. Gen.
Psychiatry 59 (5), 449–456.

Curran-Everett, D., 2000. Multiple comparisons: philosophies and illustrations. Am.

J. Physiol. Regul. Integr. Comp. Physiol. 279 (1), R1–R8.

Demjaha, A., Morgan, K., Morgan, C., Landau, S., Dean, K., Reichenberg, A., Sham, P.,

Fearon, P., Hutchinson, G., Jones, P.B., Murray, R.M., Dazzan, P., 2009. Combining di-
mensional and categorical representation of psychosis: the way forward for DSM-V
and ICD-11? Psychol. Med. 39 (12), 1943–1955.

Dikeos, D.G., Wickham, H., McDonald, C., Walshe, M., Sigmundsson, T., Bramon, E.,

Grech, A., Toulopoulou, T., Murray, R., Sham, P.C., 2006. Distribution of symptom
dimensions across Kraepelinian divisions. Br. J. Psychiatry 189 (4), 346–353.

Dixon, T., Kravariti, E., Frith, C., Murray, R.M., McGuire, P.K., 2004. Effect of symptoms

on executive function in bipolar illness. Psychol. Med. 34 (5), 811–821.

Dominguez, M.G., Viechtbauer, W., Simons, C.J., van Os, J., Krabbendam, L., 2009. Are

psychotic psychopathology and neurocognition orthogonal? A systematic review
of their associations. Psychol. Bull. 135 (1), 157–171.

Fearon, P., Kirkbride, J.B., Morgan, C., Dazzan, P., Morgan, K., Lloyd, T., Hutchinson, G.,

Tarrant, J., Fung, W.L., Holloway, J., Mallett, R., Harrison, G., Leff, J., Jones, P.B.,
Murray, R.M., 2006. Incidence of schizophrenia and other psychoses in ethnic mi-
nority groups: results from the MRC AESOP Study. Psychol. Med. 36 (11),
1541–1550.

Glahn, D.C., Almasy, L., Barguil, M., Hare, E., Peralta, J.M., Kent Jr., J.W., Dassori, A.,

Contreras, J., Pacheco, A., Lanzagorta, N., Nicolini, H., Raventos, H., Escamilla, M.A.,
2010. Neurocognitive endophenotypes for bipolar disorder identified in multiplex
multigenerational families. Arch. Gen. Psychiatry 67 (2), 168–177.

Gold, J.M., Carpenter, C., Randolph, C., Goldberg, T.E., Weinberger, D.R., 1997. Auditory

working memory and Wisconsin Card Sorting Test performance in schizophrenia.
Arch. Gen. Psychiatry 54 (2), 159–165.

Gottesman, I.I., Gould, T.D., 2003. The endophenotype concept in psychiatry: etymolo-

gy and strategic intentions. Am. J. Psychiatry 160 (4), 636–645.

Heydebrand, G., Weiser, M., Rabinowitz, J., Hoff, A.L., DeLisi, L.E., Csernansky, J.G., 2004.

Correlates of cognitive deficits in first episode schizophrenia. Schizophr. Res. 68
(1), 1–9.

Howard, R., Rabins, P.V., Mary, V., Seeman, M.V., Jeste, D.V., International Late-Onset

Schizophrenia Group, 2000. Late-onset schizophrenia and very-late-onset
schizophrenia-like psychosis: an international consensus. Am. J. Psychiatry 157
(2), 172–178.

Jablensky, A., 2006. Subtyping schizophrenia: implications for genetic research. Mol.

Psychiatry 11 (9), 815–836.

Joyce, E.M., Roiser, J.P., 2007. Cognitive heterogeneity in schizophrenia. Curr. Opin. Psy-

chiatry 20 (3), 268–272.

Kaymaz, N., Van Os, J., 2009. Murray et al. (2004) revisited: is bipolar disorder identical

to schizophrenia without developmental impairment? Acta Psychiatr. Scand. 120
(4), 249–252.

Koenen, K.C., Moffitt, T.E., Roberts, A.L., Martin, L.T., Kubzansky, L., Harrington, H.,

Poulton, R., Caspi, A., 2009. Childhood IQ and adult mental disorders: a test of
the cognitive reserve hypothesis. Am. J. Psychiatry 166 (1), 50–57.

Kravariti, E., Dixon, T., Frith, C., Murray, R., McGuire, P., 2005. Association of symptoms

and executive function in schizophrenia and bipolar disorder. Schizophr. Res. 74
(3), 221–231.

Liddle, P., 1987. The symptoms of chronic schizophrenia. A re-examination of the

positive–negative dichotomy. Br. J. Psychiatry 151 (2), 145–151.

Lindenmayer, J.P., Bossie, C.A., Kujawa, M., Zhu, Y., Canuso, C.M., 2008. Dimensions of

psychosis in patients with bipolar mania as measured by the positive and negative
syndrome scale. Psychopathology 41 (4), 264–270.

Ling, X.B., Cohen, H., Jin, J., Lau, I., Schilling, J., 2009. FDR made easy in differential fea-

ture discovery and correlation analyses. Bioinformatics 25 (11), 1461–1462.

Lucas, S., Fitzgerald, D., Redoblado-Hodge, M.A., Anderson, J., Sanbrook, M., Harris, A.,

Brennan, J., 2004. Neuropsychological correlates of symptom profiles in first epi-
sode schizophrenia. Schizophr. Res. 71 (2–3), 323–330.

MacCabe, J.H., Lambe, M.P., Cnattingius, S., Sham, P.C., David, A.S., Reichenberg, A.,

Murray, R.M., Hultman, C.M., 2010. Excellent school performance at age 16 and
risk of adult bipolar disorder: national cohort study. Br. J. Psychiatry 196 (2),
109–115.

Mallett, R., 1997. Sociodemographic Schedule. Section of Social Psychiatry, Institute of

Psychiatry, London.

Morgan, C., Dazzan, P., Morgan, K., Jones, P., Harrison, G., Leff, J., Murray, R., Fearon, P.,

2006. First episode psychosis and ethnicity: initial findings from the AESOP study.
World Psychiatry 5 (1), 40–46.

Murray, R.M., Sham, P., Van Os, J., Zanelli, J., Cannon, M., McDonald, C., 2004. A develop-

mental model for similarities and dissimilarities between schizophrenia and bipo-
lar disorder. Schizophr. Res. 71 (2–3), 405–416.

Peralta, V., Cuesta, M.J., Farre, C., 1997. Factor structure of symptoms in functional psy-

choses. Biol. Psychiatry 42 (9), 806–815.

Raven, J.C., 1995. Coloured Progressive Matrices. Sets A, AB, B. Oxford Psychologists

Press Ltd., Oxford.

Reitan, R.M., 1958. Validity of the Trail Making Test as an indication of organic brain

damage. Percept. Mot. Skills (8), 271–276.

Simonsen, C., Sundet, K., Vaskinn, A., Birkenaes, A.B., Engh, J.A., Faerden, A., Jonsdottir,

H., Ringen, P.A., Opjordsmoen, S., Melle, I., Friis, S., Andreassen, O.A., 2009.
Neurocognitive dysfunction in bipolar and schizophrenia spectrum disorders

230

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231

Author's personal copy

depends on history of psychosis rather than diagnostic group. Schizophr. Bull. 37
(1), 73–83.

Smith, M.J., Barch, D.M., Csernansky, J.G., 2009. Bridging the gap between schizophre-

nia and psychotic mood disorders: relating neurocognitive deficits to psychopa-
thology. Schizophr. Res. 107 (1), 69–75.

Spreen, O., Strauss, E., 1991. A compendium of neuropsychological tests: administra-

tion, norms, and commentary. Oxford University Press, New York.

Squires, R.F., Saederup, E., 1991. A review of evidence for GABergic predominance/

glutamatergic deficit as a common etiological factor in both schizophrenia and
affective psychoses: more support for a continuum hypothesis of “functional”
psychosis. Neurochem. Res. 16 (10), 1099–1111.

StataCorp, 2007. STATA v.10.0 for Windows. Stata Corportation, College Station, Texas.
Storey, J.D., Tibshirani, R., 2003. Statistical significance for genomewide studies. Proc.

Natl. Acad. Sci. U. S. A. 100 (16), 9440.

Strimmer, K., 2008. A unified approach to false discovery rate estimation. BMC Bioin-

formatics 9, 303.

Stroup, T.S., 2007. Heterogeneity of treatment effects in schizophrenia. Am. J. Med. 120

(4), S26–S31.

Van der Does, A.J., Dingemans, P.M., Linszen, D.H., Nugter, M.A., Scholte, W.F., 1993.

Symptom dimensions and cognitive and social functioning in recent-onset schizo-
phrenia. Psychol. Med. 23 (3), 745–753.

Van Os, J., Gilvarry, C., Bale, R., Van Horn, E., Tattan, T., White, I., Murray, R., 1999. A

comparison of the utility of dimensional and categorical representations of psycho-
sis. UK700 Group. Psychol. Med. 29 (3), 595–606.

Wechsler, D., 1981. Wechsler Adult Intelligence Scale-Revised manual. The Psycholog-

ical Corporation, USA.

Wechsler, D., 1987. Instruction Manual for the Wechsler Memory Scale Revised. Psy-

chological Corp., New York.

World Health Organization, 1992. International statistical classification of diseases and re-

lated health problems: ICD-10, 10th revision. ed. World Health Organization, Geneva.

World Health Organization, 1994. SCAN (Schedules for Clinical Assessment in Neuro-

psychiatry). World Health Organization, Geneva.

Zanelli, J., Reichenberg, A., Morgan, K., Fearon, P., Kravariti, E., Dazzan, P., Morgan, C.,

Zanelli, C., Demjaha, A., Jones, P.B., Doody, G.A., Kapur, S., Murray, R.M., 2010. Spe-
cific and generalized neuropsychological deficits: a comparison of patients with
various first-episode psychosis presentations. Am. J. Psychiatry 167 (1), 78–85.

231

E. Kravariti et al. / Schizophrenia Research 140 (2012) 221–231