Dimensions of Psychosis in Affected Sibling Pairs
by Alastair Q. Cardno, Lisa A. Jones, Kieran C. Murphy, Robert D. Sanders,
Philip Asherson, Michael J. Owen, and Peter McQuffin
molecular genetic studies of schizophrenia and to help
resolve issues concerning genetic heterogeneity. With this
as a background, work has begun to investigate the familial basis of the dimensions. In an epidemiologically based
family study, Kendler et al. (1994) found that probands'
scores on the positive and negative dimensions did not
predict the presence of schizophrenia or spectrum disorders in the probands' relatives. Cardno et al. (1997) confirmed these results in a study of family-history data, but
found that probands' scores on two factors akin to the disorganization dimension (positive formal thought disorder
and inappropriate affect/bizarre behavior) specifically predicted the presence of nonaffective psychosis in relatives.
Van Os et al. (1997) also performed a family-historybased study but on a proband sample with a range of
functional psychoses not restricted to schizophrenia. For
the whole sample, probands' scores on the negative
dimension best predicted the presence of psychosis in relatives. When the sample was restricted to probands with
schizophrenia, scores on a dimension characterized by
inappropriate affect and bizarre behavior again predicted
the presence of psychosis in relatives (positive formal
thought disorder was not included in this analysis).
Three other studies have examined associations of
dimension scores in affected sibling pairs. In a study of 98
pairs with DSM-III-R schizophrenia, Burke et al. (1996)
obtained significant but modest correlations within pairs
for scores on all three dimensions. Kendler et al. (1997)
enlarged this sample to 256 pairs and found correlations
of similar size for the positive and negative dimensions (a
disorganization dimension was not included in the factor
solution). Loftus et al. (1996), in a sample of 53 sibling
pairs with DSM-III-R schizophrenia or schizoaffective
disorder, found a trend toward an association only for the
disorganization dimension, which became statistically significant in an enlarged sample of 114 pairs.
Abstract
Factor analytical studies of schizophrenia symptoms
have consistently suggested three or more symptom
dimensions, but it is not known whether any of these
dimensions have a genetic basis. The purpose of this
study was to investigate to what extent the dimensions
show familial aggregation. Symptom ratings were
made using the SAPS and SANS and the OPCRIT
checklist on the members of 109 sibling pairs with
DSM-IV schizophrenia or schizoaffective disorder.
Factor analyses were performed on the ratings of both
instruments, and correlations were made of withinpair factor scores. Analyses were also performed on
the 89 pairs in which both members had a diagnosis of
schizophrenia. Factor analysis of SAPS and SANS ratings resulted in positive, negative, and disorganization
factors; analysis of OPCRIT ratings resulted in positive, negative, disorganization, and first-rank delusion
factors. Only the disorganization dimension showed
significant within-pair correlations, but these were of
modest size and not significantly greater than the correlations for the other dimensions. None of the dimensions showed sufficient familial aggregation to suggest
that they are close markers of genetic or common environmental factors that contribute liability to schizophrenia. They may be weakly associated with such factors and with factors that do not contribute liability to
schizophrenia but do influence the form taken by the
illness.
Key words: Sibling pairs, psychosis.
Schizophrenia Bulletin, 25(4):841-850,1999.
It is well established that the psychotic symptoms characterizing schizophrenia may be divided into three or more
dimensions by means of factor analysis (Liddle 1987;
Murphy et al. 1994; Andreasen et al. 1995). If one or
more of the dimensions showed substantial heritability,
they could potentially be used to refine the phenotype for
Send reprint requests to Dr. A. Cardno, Department of Psychological
Medicine, University of Wales College of Medicine, Heath Park, Cardiff
CF4 4XN, Wales, U.K.; e-mail: [email protected].
841
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
A.G. Cardno et al.
The most common approach has been to base factor
analyses of psychotic symptoms on ratings from the
Scales for the Assessment of Negative and Positive
Symptoms (SANS, SAPS; Andreasen 1984a, 1984fc). The
solution of positive, negative, and disorganization dimensions has been replicated in numerous studies (Liddle
1987; Andreasen et al. 1995), has been supported by confirmatory factor analysis (Peralta et al. 1994;
Lenzenweger and Dworkin 1996), and has had stability
over time (Arndt et al. 1995). Further, the dimensions
have had some specificity with respect to neuropsychological test performance (Bilder et al. 1985; Liddle and
Morris 1991) and patterns of activity on functional brain
imaging (Liddle et al. 1992).
However, none of the above studies investigating the
familial basis of the dimensions used the SAPS and
SANS. It would therefore be valuable to investigate the
evidence of familiality in dimensions based on SAPS and
SANS ratings. It would also be useful to employ a clinical
rating scale with a different structure from the SAPS and
SANS in the same patient population. Using a rating scale
with a different structure would help to determine to what
extent the commonly found symptom dimensions are
dependent on the nature of the clinical rating scale
employed; it would also help to establish what kind of
dimension has the strongest familial basis.
The OPCRIT checklist (McGuffin et al. 1991) is a
suitable alternative clinical rating instrument. While the
SAPS and SANS define each symptom on a 0-5 ordinal
scale, OPCRIT psychotic symptoms are defined categorically. OPCRIT also includes a wider range of psychotic
symptoms, particularly positive symptoms, than the SAPS
and SANS global scales and was designed for lifetimeever symptom ratings for genetic studies. This instrument
was used in the family-history study of psychotic symptom dimensions mentioned above (Cardno et al. 1997),
and it has been used in other factor-analytical studies of
psychotic symptoms (Cardno et al. 1996a, 1996b) and the
above family-history study that combined symptoms and
demographic variables (van Os et al. 1997). Because of
the larger number of positive symptoms in OPCRIT, factor solutions have often involved division of the positive
factor to include first-rank symptom factors (Cardno et al.
1996a, 1997).
The purposes of this study were to perform factor
analyses of SAPS and SANS global ratings and OPCRIT
psychotic symptoms, and to investigate associations of
factor scores within sibling pairs with schizophrenia or
schizoaffective disorder.
Methods
Patients. Families in which two or more siblings met
the criteria for DSM-TV schizophrenia or schizo-affective
disorder (American Psychiatric Association 1994) were
ascertained through mental health services and relatives'
support groups in England and Wales. After subjects were
given a complete description of the study, written
informed consent was obtained. The sample comprised
191 individuals, grouped as 109 sibling pairs (82 pairs; 9
trios, each counting as 3 pairs). The sex distribution was
133 males and 58 females. All individuals were Caucasian
and had been born in the United Kingdom. Mean age was
42.0 years (standard deviation (SD) 12.3) and mean age at
illness onset was 24.4 years (SD 7.6). In 89 pairs both
members had a diagnosis of schizophrenia, in 4 pairs both
had a diagnosis of schizoaffective disorder, and in the
remaining 16 pairs one member had each diagnosis.
Diagnoses were based on all available clinical information including a semi-structured interview (Present
State Examination-9, Wing et al. 1974; or Schedules for
Clinical Assessment in Neuropsychiatry, Wing et al.
1990), examination of case records, and information from
relatives and mental health professionals. The data were
compiled into case vignettes for each individual.
Diagnoses were made independently by the interviewer
and another member of the diagnostic team (A.G.C.,
L.A.J., K.C.M.). If they agreed on a diagnosis it became
the consensus diagnosis. In cases of disagreement the
third member of the team made a independent diagnosis,
and the case was discussed in detail in order to reach consensus. Diagnostic reliability was assessed on 40 cases
with a range of functional psychoses. Diagnoses made
independently by each team member were compared with
consensus diagnoses. Mean kappa score was 0.89.
Symptom Ratings. Research interviews and ratings of
clinical variables for the individual members of each sibling pair were carried out by different members of the
clinical assessment team, except where this was not practicable (for example, for geographical reasons). Individual
members of pairs were rated independently in 72 percent
of cases. The versions of SAPS and SANS incorporated
into the Diagnostic Interview for Genetic Studies
(Nurnberger et al. 1994) were completed for each individual, based on all available clinical information. The clinical assessment team (A.G.C., L.A.J., K.C.M., R.D.S.)
established rating guidelines to supplement the symptom
definitions accompanying the scales; some rating criteria
842
Dimensions of Psychosis in Affected Sibling Pairs
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
were modified in minor ways to facilitate the rating of
symptoms at their most severe on a lifetime-ever basis
and to minimize the risk of false positive ratings. The
global ratings for avolition-apathy, anhedonia-asociality,
and attention were not included in the analysis because
we were not confident that these could be rated accurately
on a lifetime-ever basis from the information available to
us. Inappropriate affect and poverty of content of speech
were considered separately from the other symptoms in
their respective sections, because previous factor analyses
have suggested that these symptoms form part of the disorganization dimension (Andreasen et al. 1995) and therefore should not contribute to global ratings of negative
symptoms. The OPCRIT checklist was also completed for
each individual on the basis of all available clinical information. Similar rating guidelines were established to supplement the symptom definitions accompanying the
checklist. In addition, the Global Assessment Scale (GAS;
Endicott et al. 1976) was completed for each individual as
a general measure of illness severity.
Reliability of ratings for SAPS-SANS, OPCRIT, and
GAS was assessed on 30 cases with a range of functional
psychoses. The ratings of each member of the clinical
assessment team were compared with consensus ratings.
Ideally, the study would have had separate teams making
independent consensus ratings of the members of each
sibling pair, so the consensus ratings were our clinical
"gold standard." Each member's ratings were compared
with the consensus ratings rather than simply with the ratings of colleagues. For quantitative data (SAPS-SANS,
age at onset, and GAS), reliability was expressed as the
mean intraclass correlation between each rater and the
consensus ratings. For OPCRIT symptoms, mean kappa
scores were employed. In order to optimize clinical rating
standards, weekly meetings were held throughout the
duration of the study in which members of the clinical
assessment team made appropriate ratings of prepared
case vignettes or audiotape or videotape recordings of
interviews.
was determined by the "eigenvalue greater than 1" rule,
and these were subjected to both oblique (direct oblimin)
and orthogonal (varimax) rotations. Correlations between
factors following oblique rotation were examined to
determine the most appropriate form of rotation.
Regression factor scores were then based on the factor
loadings of that rotation. For the purposes of description,
high factor loadings were considered to be those greater
than 0.50.
For the factor analysis of OPCRIT symptoms, the 26
single psychotic symptom items were initially considered.
Analysis then proceeded in the same way as for the
SAPS-SANS ratings, except that the scree test (Cattell
1966) was used to determine the number of substantive
factors. We have found from previous factor analyses of
OPCRIT symptoms that the "eigenvalue greater than 1"
rule tends to produce an excessively large number of factors, often characterized by only one symptom. In addition to analyses of the whole sample of 191 individuals,
analyses were performed of SAPS-SANS and OPCRIT
ratings for the 157 individuals from the 89 sibling pairs
where both members had a diagnosis of schizophrenia.
The symptom ratings cannot be regarded as independent observations because the sample comprised pairs and
trios of siblings. However, although this may affect standard errors of estimates, it should not substantively affect
the factor structure (Murphy et al. 1994). This is supported by the similar results found for exploratory factor
analysis using unrelated (Liddle 1987; Andreasen et al.
1995) and related individuals (Murphy et al. 1994), and
for confirmatory factor analysis using unrelated (Peralta et
al. 1994) and related individuals (Lenzenweger and
Dworkin 1996).
Spearman correlations were used to compare factor
scores based on SAPS-SANS and OPCRIT ratings, and
also to perform within-pair correlations of factor scores
for each dimension. The treatment of sibling trios as three
pairs raises the issue of the correlation for the "third" pair
being dependent on that of the other two. In sibships of
only three, the effect of this is likely to be minimal, but as
a check we also performed within-pair correlations of the
91 sibships, randomly taking one pair from each trio.
Within-pair correlations were hypothesized to be the
result of common underlying factors; therefore, the correlation coefficient could be regarded as an estimate of the
variance accounted for by the association. The size of correlation coefficients among the different dimensions were
compared using the approach outlined in the appendix.
Where appropriate, the potential confounding effects of
sex, age at onset, illness duration, and illness severity
(measured as worst-ever GAS score) were controlled for
using partial correlations. The reliability correlation coefficient for age at onset was 0.99 and for illness severity it
Statistics. Statistical analyses were performed using
SPSS for Windows, Version 6.0. For the factor analysis of
SAPS-SANS, the following items were initially considered: inappropriate affect, affective flattening, poverty of
content of speech, alogia, hallucinations, delusions,
bizarre behavior, and positive formal thought disorder.
Any item rated as definitely present (>1) in less than 10
percent of cases was then excluded from the analysis, as
was any item with a reliability correlation coefficient of
less than 0.41 (that is, the lower limit of the "moderate
agreement" range suggested by Landis and Koch (1977)).
Those items that remained underwent a principle components analysis. The number of substantive factors present
843
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
A.G. Cardno et al.
symptoms (table 1). The results of the factor analysis of
SAPS-SANS ratings for the 157 individuals from sibling
pairs where both members had a diagnosis of schizophrenia were similar: Factor 1 (disorganization) had high loadings for inappropriate affect (0.78), bizarre behavior
(0.77), and positive formal thought disorder (0.63); Factor
2 (negative) had high loadings for affective flattening
(0.86) and alogia (0.84); and Factor 3 (positive) had high
loadings for hallucinations (0.84) and delusions (0.79).
For the factor analysis of OPCRIT ratings on the
whole sample, the following symptoms were excluded
because they were rated as present in less than 10 percent
of cases: incoherent speech, primary delusional perception, thought withdrawal, and thought echo. The remaining 22 symptoms had sufficiently high reliability kappa
scores to be retained for the factor analysis. The lowest
scoring symptoms were bizarre behavior (0.52), blunted
affect (0.53), and other primary delusions (i.e., other than
delusional perception) (0.54). The other symptoms had
kappa scores ranging from 0.64 to 0.92. Factor analysis
suggested four substantive factors that accounted for 38.8
percent of the variance. Correlations between the factors
following oblique rotation were again all low (-0.21 to
0.12), so further analysis was based on factor loadings
following orthogonal rotation. The factors were characterized respectively by positive, disorganization, and negative symptoms, and first-rank delusions (table 2). Factor
analysis of OPCRIT ratings for the 157 individuals in
which both members of the sibling pair had a diagnosis of
schizophrenia produced similar results, although somewhat changed in size order: Factor 1 (positive) had high
loadings for third-person voices (0.58), commentary
voices (0.52), abusive voices (0.55), and other hallucinations (0.61); the loading for delusions of influence
dropped slightly to 0.43. Factor 2 (first-rank delusions)
had high loadings for persecutory delusions (0.52—up
from 0.35), bizarre delusions (0.63), delusions of passiv-
was 0.77. All tests of statistical significance were twotailed.
In case the method of using factor scores should mask
familial aggregation of dimensions, a further analysis was
performed using symptom scores instead of factor scores.
Based on the results of previous confirmatory factor analyses, we summed the scores for the characteristic symptoms
of each dimension (and symptom scores for OPCRTT firstrank symptom dimensions). The following symptoms were
included: delusions and hallucinations for the positive
dimension; affective flattening and alogia for the negative
dimension; positive formal thought disorder and inappropriate affect for the disorganization dimension; delusions
of passivity, thought insertion, and broadcast for the firstrank delusions dimension; and third-person voices and
commentary voices for the first-rank hallucinations dimension. Again, Spearman correlations were used to examine
within-pair association of dimensions.
Results
Factor Analyses. For the factor analysis of SAPSSANS ratings on the whole sample of 191 individuals,
poverty of content of speech was rated as definitely present in less than 10 percent of cases and so was excluded
from the analysis. All the remaining items had sufficiently
high reliability correlation coefficients to be retained for
the factor analysis. The lowest scoring items were bizarre
behavior (0.65) and alogia (0.68); correlations for the
other items ranged from 0.72 to 0.93. The factor analysis
suggested three substantive factors, which accounted for
65.5 percent of the variance. Correlations between factors
following oblique rotation were all low (-0.15 to 0.07), so
further analysis employed factor loadings following
orthogonal rotation. The factors were characterized
respectively by disorganization, and negative and positive
Table 1. Factor analysis of SAPS-SANS ratings for the whole sample (N = 191)
Symptom
Factor 1
Factor 2
Inappropriate affect
Affective flattening
Alogia
Hallucinations
Delusions
Bizarre behavior
Factor 3
Positive formal thought disorder
0.78
0.10
0.07
-0.02
0.10
0.73
0.69
0.15
0.86
0.85
0.00
-0.14
0.15
-0.06
-0.13
0.00
-0.14
0.84
0.81
0.13
0.07
Variance accounted for, %
27.5
22.6
15.5
Predominant symptoms
Note.—Loadings of > 0.50 in bold.
Disorganization
844
Negative
Positive
Dimensions of Psychosis in Affected Sibling Pairs
Table 2.
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
Factor analysis of OPCRIT ratings for the whole sample {N = 191)
Symptom
Factor 1
Factor 2
Factor 3
Factor 4
Bizarre behavior
Catatonia
Speech difficult to understand
Positive formal thought disorder
Negative formal thought disorder
Restricted affect
Blunted affect
Inappropriate affect
Persecutory delusions
Grandiose delusions
Delusions of influence
Bizarre delusions
Delusions of passivity
Other primary delusions
Thought insertion
Thought broadcast
Nihilistic delusions
Third-person voices
Commentary voices
Abusive voices
Other auditory hallucinations
Other hallucinations
Variance accounted for, %
-0.10
-0.21
-0.03
-0.01
-0.32
0.17
-0.06
-0.19
0.46
0.15
0.53
0.06
0.05
0.30
0.07
0.05
0.20
0.61
0.55
0.69
0.35
0.57
14.8
0.18
0.17
0.78
0.80
-0.11
-0.03
-0.01
0.46
0.02
0.51
0.13
0.27
-0.07
0.31
0.00
-0.05
0.26
-0.13
-0.17
-0.07
0.21
0.11
9.9
0.27
0.52
0.08
0.05
0.64
0.68
0.54
0.24
-0.24
-0.15
-0.18
-0.19
0.06
-0.07
0.19
0.17
-0.27
0.04
-0.03
-0.16
0.26
-0.14
7.4
0.03
-0.14
-0.14
0.00
0.05
-0.01
-0.09
0.00
0.35
0.20
0.14
0.65
0.76
0.20
0.60
0.41
-0.10
0.19
-0.18
0.11
-0.12
0.04
6.6
Predominant symptoms
Positive
Disorganization
Negative
First-rank
Delusions
Note.—Loadings of > 0.50 in bold.
ity (0.73), and thought insertion (0.62). Factor 3 (disorganization) had high loadings for speech difficult to understand (0.76), positive formal thought disorder (0.78), and
inappropriate affect (0.53—up from 0.46); grandiose
delusions dropped slightly to 0.49. Factor 4 (negative) had
high loadings for negative formal thought disorder (0.50),
restricted affect (0.66), and blunted affect (0.54); catatonia
dropped slightly to 0.48.
The correlations between SAPS-SANS and OPCRIT
factor scores for the whole sample were as follows: positive dimension, r = 0.67, p < 0.0005; negative dimension,
r = 0.74, p < 0.0005; and disorganization dimension r =
0.62, p < 0.0005. The OPCRIT first-rank delusions
dimension had its highest correlation with the SAPSSANS positive dimension (r = 0.29, p < 0.0005).
Correlations of Factor Scores Within Sibling Pairs.
These correlations are shown in table 3. For the SAPSSANS analysis on the whole sample, only factor scores
for the disorganization dimension were significantly correlated within sibling pairs (r = 0.20, p = 0.04) at an
uncorrected alpha level of 5 percent. For the schizophrenia-only sibling pairs, again only the disorganization
dimension showed significant within-pair correlation (r =
845
0.22, p = 0.03). However, if a Bonferroni correction for
multiple testing is applied, neither of these correlations
remains, significant. In addition, the correlation for the
disorganization dimension was not significantly greater
than those for the other dimensions in either the whole or
schizophrenia-only sample.
For the OPCRIT analysis on the whole sample, once
again only scores for the disorganization dimension
showed a significant uncorrected within-pair correlation
(r = 0.32, p = 0.001). In view of the fact that grandiose
delusions had a higher than expected factor loading on the
disorganization dimension, we repeated the OPCRIT
factor analysis omitting this symptom. This resulted in
similar positive, disorganization, negative, and first-rank
delusions dimensions. The highest loadings on the disorganization factor were for speech difficult to understand
(0.80), positive formal thought disorder (0.79), and inappropriate affect (0.50). The within-pair correlation for the
disorganization dimension was r = 0.33, p - 0.001.
Within-pair correlations for the other dimensions were
positive (r = 0.16); negative (r = 0.14); and first-rank
delusions (r = 0.10). For the schizophrenia-only sibling
pairs (including grandiose delusions), the disorganization
dimension remained significantly correlated (r = 0.32, p =
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
Table 3.
A.G. Cardno et al.
Correlations of factor scores within sibling pairs
All pairs (W = 109)
Rating
instrument
Correlation
coefficient (r)
SAPS-SANS
Positive
Negative
Disorganization
0.11
0.11
0.20
Significance
(P)1
0.25
0.26
0.04
OPCRIT
Positive
Negative
Disorganization
First-rank delusions
0.18
0.16
0.32
0.11
0.07
0.09
0.001
0.25
Dimension
Schizophrenia pairs (n = 89)
Correlation
coefficient (r)
0.13
0.17
0.22
0.23
0.12
0.03
0.21
0.16
0.32
0.07
0.047
0.14
0.002
0.51
Significance
(P)1
1
Uncorrected p values statistically significant if < 0.05; after Bonferroni correction for multiple testing SAPS-SANS correlations significant
if p < 0.0167, and OPCRIT correlations significant if p < 0.0125.
0.002), and the positive dimension also showed significant within-pair correlation (r = 0.21, p = 0.047). The correlations for the disorganization dimension would remain
significant after correction for multiple testing, but that
for the positive dimension in the schizophrenia-only sample would become nonsignificant. The correlations for the
disorganization dimension were not significantly greater
than those for the other dimensions, nor were they significantly greater than the correlations for the disorganization
dimension derived from SAPS-SANS ratings.
Within-pair correlations for the 91 pairs from separate sibships were, for SAPS-SANS-derived factor scores,
positive dimension, r = 0.17, p = 0.11; negative dimension, r - 0.18, p - 0.10; and disorganization dimension,
r = 0.20, p = 0.06. For OPCRIT-derived scores the correlations were positive dimension, r = 0.19, p = 0.07; negative dimension, r = 0.17, p = 0.11; disorganization dimension, r = 0.36, p = 0.001; and first-rank delusions
dimension, r = 0.14, p = 0.18.
Partial correlations of SAPS-SANS disorganization
factor scores in the whole sample, controlling for potential confounding variables were as follows: controlling for
sex, r - 0.17, p = 0.07; for age at onset, r = 0.15, p = 0.13;
for illness duration, r - 0.17, p = 0.09; for illness severity,
r = 0.12, p = 0.22; and for all four variables combined, r =
0.14, p = 0.17. The partial correlations of OPCRIT disorganization factor scores were as follows: controlling for
sex, r = 0.32, p = 0.001; for age at onset, r = 0.31, p =
0.001; for illness duration, r - 0.30, p = 0.003; for illness
severity, r = 0.25, p = 0.01; and for all four variables combined, r = 0.22, p = 0.03.
Within-pair correlations of SAPS and SANS symptom scores in the whole sample were as follows: positive
dimension, r = 0.11, p = 0.25; negative dimension, r =
0.15, p - 0.12; and disorganization dimension, r = 0.20,
p = 0.03. The following were the correlations for OPCRIT
symptom scores: positive dimension, r = 0.21, p = 0.03;
846
first-rank delusions dimension, r = 0.00, p = 0.98; firstrank hallucinations dimension, r = -0.03, p = 0.75; negative dimension, r = 0.12, p = 0.21; and disorganization
dimension, r = 0.25, p = 0.01.
Discussion
Factor Analyses. The factor analysis of SAPS and
SANS ratings for the whole sample of 191 individuals
resulted in positive, negative, and disorganization factors
that were similar to those commonly found in other factor-analytical studies of psychotic symptoms (Liddle
1987; Murphy et al. 1994; Andreasen et al. 1995). The
factor analysis, which was limited to the 157 individuals
from sibling pairs in which both members had a diagnosis
of schizophrenia, resulted in the same three dimensions
and similar factor loadings for the characteristic symptoms of each dimension.
The factor analysis of OPCRIT ratings for the whole
sample resulted in four dimensions, three of which were
similar to those found in the SAPS-SANS analyses; that
is, positive, negative, and disorganization factors, and a
fourth factor that had high loadings for first-rank and
bizarre delusions. The positive dimension had moderate
or high loadings for a range of delusions and hallucinations; however, the high loadings were most commonly
associated with hallucinations. The negative dimension
was characterized by poverty of speech and restricted
affect, as is commonly found, and also had a moderately
high loading for catatonia, possibly because this OPCRIT
item includes stupor. The disorganization dimension was
characterized by positive formal thought disorder, as
expected. There was also a moderate loading for inappropriate affect (0.46), which increased slightly to 0.53 in the
analysis of individuals from schizophrenia-only sibling
pairs. However, there was an unexpectedly high loading
Dimensions of Psychosis in Affected Sibling Pairs
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
of 0.51 for grandiose delusions. This result was not
explained by the presence of individuals with schizoaffective disorder because the loading dropped only slightly (to
0.49) in the schizophrenia-only analysis. In support of the
view that grandiose delusions do not play a major role in
characterizing the disorganization dimension, the factor
solution from the analysis in which grandiose delusions
were omitted was similar to the solution with this item
included, and correlations of factor scores within sibling
pairs also changed little when grandiose delusions were
not included in the analysis.
The fourth dimension, characterized by first-rank and
bizarre delusions, probably resulted from the larger number of positive symptoms included in the OPCRIT factor
analysis compared with the SAPS-SANS analysis, which
included only global ratings of hallucinations and delusions. This class of delusions and experiences is a recognizable clinical entity and formed a distinct dimension in
the two previous factor analyses of OPCRIT symptoms
that had solutions of greater than three dimensions
(Cardno et al. 1996a, 1997). Other factor analytical studies have also found first-rank symptoms to load separately
from other positive symptoms (Liddle 1987; Jorgensen
and Parnas 1990). The factor analysis of OPCRIT ratings
from the individuals who were members of schizophrenia-only sibling pairs produced generally similar results,
although not as similar as the two corresponding SAPSSANS analyses.
The solution of the SAPS-SANS whole sample
analysis accounted for a larger proportion of the variance
than the OPCRIT analysis, and had a simpler structure,
but there were substantial correlations among the three
SAPS-SANS dimensions and their OPCRIT counterparts
(r = 0.62-O.74), suggesting that these dimensions are for
the most part not scale-specific.
significantly greater than the correlations for the other
dimensions. The correlation coefficient (0.32) and variance explained was higher than for SAPS-SANS, but not
significantly so, and it was still of modest size. The correlation was not accounted for by the unexpectedly high
loading for grandiose delusions, because analysis without
this symptom produced similar results. The correlation
within the schizophrenia-only sample of sibling pairs was
also similar. In this sample, the correlation for the positive
dimension also became significant, but it would not
remain so if corrected for multiple testing. The correlation
coefficient for disorganization in the whole sample was
reduced when controlling for the potential confounding
variables (illness severity having the largest effect).
However, it remained significant, suggesting that the
familial aggregation of disorganization dimension scores
could not be accounted for solely by these factors, at least
as they were measured in this study.
Within-pair correlations for the 91 pairs from separate sibships were the same or slightly greater than those
for all 109 pairs. There was therefore no evidence that
counting trios as three pairs artificially elevated correlations. Within-pair correlations of symptom scores were
similar to the correlations of factor scores for the three
main dimensions, and zero for first-rank symptom dimensions. Symptom scores do not appear to define more
familial dimensions than factor scores, but they may be a
practical alternative for the three-dimension model when
ease of calculation is important.
The finding of a relatively weak within-pair association for scores on the disorganization dimension is consistent with the results of the two previous studies of psychotic dimensions in affected sibling pairs, which
included analysis of this dimension (Burke et al. 1996;
Loftus et al. 1996). Limiting the association to the disorganization dimension is consistent with one of these studies (Loftus et al. 1996) despite the fact that the study used
a different clinical rating instrument (Krawieka et al.
1977) and method of analysis. The dimensions were rated
as present or absent on the basis of symptom ratings, and
chi-square tests were used to determine associations
within pairs.
The two other studies of psychotic symptom dimensions were methodologically more similar to the present
one. Burke et al. (1996) found a significant correlation for
the disorganization dimension of similar magnitude to that
found in this study (raw r - 0.30), but they also found significant correlations for the positive and negative dimensions (raw r = 0.33 and 0.23, respectively). Kendler et al.
(1997), using an enlarged sample including that of Burke
et al. (1996), also found significant correlations for the
positive (r = 0.16) and negative (r = 0.21) dimensions, but
once again the correlations were small. Their factor solu-
Correlations of Factor Scores Within Sibling Pairs.
For the SAPS-SANS whole sample analysis, the only significant correlation within sibling pairs was for the disorganization dimension. However, the correlation coefficient (0.20), and hence the variance accounted for, was
small, and the correlation would become nonsignificant if
corrected for multiple testing. It was also not significantly
greater than correlations for the other dimensions. The
results for the analysis of the schizophrenia-only sibling
pairs were similar. The within-pair correlation for the
whole sample also became nonsignificant when controlling for the potential confounding variables of sex, age at
onset, illness duration, and illness severity.
For the OPCRIT whole sample analysis, again the
disorganization dimension alone was significantly correlated within sibling pairs. In this case, it would remain so
if corrected for multiple testing, although again it was not
847
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
A.G. Cardno et al.
tion did not include a disorganization dimension, probably
because positive formal thought disorder was the only
symptom characteristic of that dimension included in their
analysis. These studies differ from the present one in
terms of which within-pair correlations are statistically
significant, but the correlation coefficients for each
dimension do not differ significantly among the three
studies. Therefore, it seems probable, based on the work
carried out to date, that each of the three commonly found
psychotic symptom dimensions shows a small degree of
familial aggregation.
The modest homotypia found for the dimensions
could be a result of several mechanisms. It may be a
purely environmental or purely genetic effect, or could
result from a combination of these factors. Further work,
for example using data from twins, would help to differentiate these possibilities. The within-pair correlations are
probably too low for any of the dimensions to be close
markers of genetic or common environmental factors that
contribute liability to schizophrenia. If such factors underlay the liability to a dimension, we would expect the
within-pair correlations to be considerably higher,
because pairs of ill siblings would share most familial etiological factors.
However, it is possible that one or more of the dimensions is associated with nonshared environmental factors
contributing liability to schizophrenia, because this would
not be related to familial aggregation of dimension scores.
Dimensions may also be weakly associated with genetic
or common environmental factors contributing liability to
schizophrenia and with familial factors that influence the
form in which schizophrenia is expressed but that are not
etiological risk factors for schizophrenia. Investigation of
associations of dimension scores in twins and in other
pairs of relatives could help to investigate these possibilities further.
where zx and z2 are the z scores for the first and second
correlation coefficients, nl and n2 are the population sizes
from which the correlations were derived (in this study,
nl = n2 = the number of sibling pairs), and T represents a
range of approximately normally distributed values with
mean - 0 and variance = 1. A significant difference in correlation coefficients at an alpha level of 5 percent is indicated by a T-value greater than 1.96 or less than -1.96.
References
American Psychiatric Association. DSM-IV: Diagnostic
and Statistical Manual of Mental Disorders. 4th ed.
Washington, DC: The Association, 1994.
Andreasen, N.C. The Scale for the Assessment of Negative
Symptoms (SANS). Iowa City, I A: University of Iowa,
1984a.
Andreasen, N.C. The Scale for the Assessment of Positive
Symptoms (SAPS). Iowa City, IA: University of Iowa,
1984ft.
Andreasen, N . C ; Arndt, S.; Alliger, R.; Miller, D.; and
Flaum, M. Symptoms of schizophrenia: Methods, meanings, and mechanisms. Archives of General Psychiatry,
52:341-351, 1995.
Arndt, S.; Andreasen, N.C; Flaum, M.; Miller, D.; and
Nopoulos, P. A longitudinal study of symptom dimensions: Prediction and patterns of change. Archives of
General Psychiatry, 52:352-360, 1995.
Bilder, R.M.; Mukherjee, S.; Rieder, R.O.; and
Pandurangi, A.K. Symptomatic and neuropsychological
components of deficit states. Schizophrenia Bulletin,
ll(3):409-419, 1985.
Burke, J.G.; Murphy, B.M.; Bray, J.C; Walsh, D.; and
Kendler, K.S. Clinical similarities in siblings with schizophrenia. American Journal of Medical Genetics (Neuropsychiatric Genetics), 67:239-243, 1996.
Appendix: Comparing Correlation
Coefficients
Cardno, A.G.; Holmans, P.A.; Harvey, I.; Williams, M.B.;
Owen, M.J.; and McGuffin, P. Factor-derived subsyndromes of schizophrenia and familial morbid risks.
Schizophrenia Research, 23:231-238, 1997.
Correlation coefficients are first converted to z-scores by
means of the formula:
Cardno, A.G.; Jones, L.A.; Murphy, K.C; Asherson, P.;
Scott, L.C.; Williams, J.; Owen, M.J.; and McGuffin, P.
Factor analysis of schizophrenic symptoms using the
OPCRIT checklist. Schizophrenia Research, 22:233-239,
1996a.
\+r
z = 1/2 log.
\-r
The z-scores are then compared by means of the formula:
zx-z2
T =
Cardno, A.G.; Murphy, K . C ; Jones, L.A.; Guy, C.A.;
Asherson, P.; Azevedo, M.H.P. de.; Cruz Coelho, I.M.O.
da.; Macedo e Santos, A.J.F. de.; Pato, C.N.; McGuffin,
1
-3
848
Dimensions of Psychosis in Affected Sibling Pairs
Schizophrenia Bulletin, Vol. 25, No. 4,1999
P.; Owen, M.J.; and O'Donovan, M.C. Expanded
CAG/CTG repeats in schizophrenia: A study of clinical
correlates. British Journal of Psychiatry, 169:766-771,
1996ft.
McGuffin, P.; Farmer, A.E.; and Harvey, I.H. A polydiagnostic application of operational criteria in studies of psychotic illness: Development and reliability of the OPCRIT
system. Archives of General Psychiatry, 48:764-770,
1991.
Cattell, R.B. The screen test for the number of factors.
Multivariate Behavioral Research, 1:245-276, 1966.
Murphy, B.M.; Burke, J.G.; Bray, J.C.; Walsh, D.; and
Kendler, K.S. An analysis of the clinical features of familial schizophrenia. Acta Psychiatrica
Scandinavica,
89:421-427, 1994.
Endicott, J.; Spitzer, R.L.; Fleiss, J.L.; and Cohen, J. The
global assessment scale: A procedure for measuring overall severity of psychiatric disturbance. Archives of
General Psychiatry, 33:766-771, 1976.
Nurnberger, J.I., Jr.; Blehar, M.C; Kaufmann, C.A.; YorkCooler, C ; Simpson, S.G.; Harkavy-Friedman, J.; Severe,
J.B.; Malaspina, D.; Reich, T.; and collaborators.
Diagnostic interview for genetic studies: Rationale,
unique features, and training. Archives of General
Psychiatry, 51:849-859, 1994.
Jorgensen, A., and Parnas, J. The Copenhagen high-risk
study: Premorbid and clinical dimensions of maternal
schizophrenia. Journal of Nervous and Mental Disease,
178:370-376, 1990.
Kendler, K.S.; Karkowski-Shuman, L.; O'Neill, F.A.;
Straub, R.E.; MacLean, C.J.; and Walsh, D. Resemblance
of psychotic symptoms and syndromes in affected sibling
pairs from the Irish study of high-density schizophrenia
families: Evidence for possible etiologic heterogeneity.
American Journal of Psychiatry, 154:191-198, 1997.
Peralta, V.; Cuesta, M.J.; and de Leon, J. An empirical
analysis of latent structures underlying schizophrenic
symptoms: A four-syndrome model.
Biological
Psychiatry, 36:726-736,1994.
van Os, J.; Marcelis, M.; Sham, P.; Jones, P.; Gilvarry, K.;
and Murray, R. Psychopathological syndromes and familial morbid risk of psychosis. British Journal of
Psychiatry, 170:241-246, 1997.
Kendler, K.S.; McGuire, M.; Gruenberg, A.M.; and
Walsh, D. Clinical heterogeneity in schizophrenia and the
pattern of psychopathology in relatives: Results from an
epidemiologically based family study. Ada Psychiatrica
Scandinavica, 89:294-300, 1994.
Wing, J.K.; Babor, M.D.; Brugha, T; Burke, J.; Cooper,
J.E.; Giel, R.; Jablenski, A.; Regier, D.; and Sartorius, N.
SCAN: Schedules for clinical assessment in neuropsychiatry. Archives of General Psychiatry, 47:589-593,1990.
Krawieka, M.; Goldberg, D.; and Vaughan, M. A standardized psychiatric assessment scale for rating chronic
psychotic patients. Ada Psychiatrica
Scandinavica,
55:299-308, 1977.
Wing, J.K.; Cooper, J.E.; and Sartorius, N. The Measurement and Classification of Psychiatric
Symptoms.
Cambridge, England: Cambridge University Press, 1974.
Landis, J.R., and Koch, G.C. The measurement of
observer agreement for categorical data. Biometrics,
33:1089-1091, 1977.
Acknowledgments
Lenzenweger, M.F., and Dworkin, R.H. The dimensions
of schizophrenia phenomenology: Not one or two, at least
three, perhaps four. British Journal of Psychiatry,
168:432-440, 1996.
This work was supported by the Medical Research
Council (U.K.), and Glaxo Research and Development.
The authors thank the psychiatrists, mental health professionals, patients, and relatives who helped with this study;
Dr. Peter Holmans for statistical advice, and Professor
Kenneth Kendler for valuable comments.
Liddle, P.F. The symptoms of chronic schizophrenia: A reexamination of the positive-negative dichotomy. British
Journal of Psychiatry, 151:145-151, 1987.
Liddle, P.F.; Friston, K.J.; Frith, CD.; Hirsch, S.R.; Jones,
T; and Frackowiak, R.S.J. Patterns of cerebral blood flow
in schizophrenia. British Journal of Psychiatry,
160:179-186, 1992.
The Authors
Alastair G. Cardno, M.R.C.Psych., and Kieran C.
Murphy, M.R.C.Psych., are Medical Research Council
(U.K.) Clinical Training Fellows, Department of
Psychological Medicine; Lisa A. Jones, B.Sc, and Robert
D. Sanders, B . S c , are Research Psychologists,
Department of Psychological Medicine, University of
Wales College of Medicine (U.W.C.M.), Cardiff, U.K.
Philip Asherson, M.R.C.Psych, is Senior Lecturer in
Liddle, P.F., and Morris, D. Schizophrenic syndromes and
frontal lobe performance. British Journal of Psychiatry,
158:340-345, 1991.
Loftus, J.B.; Bass, N.J.; DeLisi, L.E.; and Crow, T.J.
Syndrome structure in schizophrenia is independent of
genetic predisposition, [abstract].
Schizophrenia
Research, 18:171, 1996.
849
Schizophrenia Bulletin, Vol. 25, No. 4, 1999
A.G. Cardno et al.
Psychiatry, Institute of Psychiatry, London. Michael J.
Owen, M.R.C.Psych., Ph.D., is Professor of Neuropsychiatric Genetics, Departments of Psychological
Medicine and Medical Genetics; Peter McGuffin,
F.R.C.Psych., Ph.D., is Professor and Head of Division,
Department of Psychological Medicine, U.W.C.M.,
Cardiff, U.K.
850