Obsessive–compulsive disorder, tics and anxiety
in 6-year-old twins
DEREK BOLTON*, FRU¨HLING RIJSDIJK, THOMAS G. O’CONNOR,
SEAN PERRIN AND THALIA C. ELEY
Institute of Psychiatry, Kings College London, UK
Background. Previous reports of genetic influences on obsessive–compulsive disorder (OCD)
symptoms have suggested moderate heritability. Family history studies of co-morbidity have found
familial aggregation with tics, especially for early-onset OCD, and familial aggregation with anxiety
Method. Heritability of OCD and familial aggregation of OCD, tics and anxiety disorders were
investigated in a community sample of 6-year-old twins using a two-phase design in which 4662
twin pairs were sampled and 854 pairs were assessed in the second phase by maternal-informant
diagnostic interview using DSM-IV criteria.
Results. In the multivariate model combined additive genetic and common environmental effects
were estimated as 47% for sub-threshold OCD, and the model was unable to distinguish these
sources of familial aggregation. There were strong familial aggregations between sub-threshold
OCD and tics and between sub-threshold OCD and other anxiety disorders (80% and 97%
respectively), although again specific sources could not be distinguished.
Conclusions. The findings are consistent with the hypothesis of a tic-related early-onset OCD
phenotype, but also with the hypothesis of an anxiety-related early-onset OCD phenotype.
The main aim of the present study was to
investigate genetic and environmental influences
on paediatric obsessive–compulsive disorder
(OCD) and on its associations with tics and with
anxiety disorders. Key points in the literature –
recently reviewed in Hettema et al. (2001),
Shih et al. (2004) and van Grootheest et al.
(2005) – include the following.
Large-scale twin studies of OCD symptoms
are rare, and there have been no studies of a
diagnostically defined phenotype. Clifford et al.
(1984) gave the Leyton Obsessional Inventory
to 419 pairs of non-clinic referred twins, esti-
mating heritability for its trait and symptom
scales as 47% and 44% respectively. More
recently Jonnal and colleagues (2000) reported a
sample of 527 adult female twin pairs who
completed 20 items from the Padua Inventory of
obsessive compulsive symptoms and estimated
heritability of the two main factors, corre-
sponding roughly to obsessions and compul-
sions; the best-fit model suggested additive
genetic heritabilities of 33% and 26% respect-
ively. Hudziak and colleagues (2004) assessed
large samples of twins aged 7, 10 and 12 years in
cohorts from Holland and the USA using an
8-item Obsessive–Compulsive Scale contained
in the Child Behavior Checklist. Best-fitting
models indicated significant additive genetic
influences in the range 45–58%, with significant
shared environmental influence detected only in
the Dutch 12-year-old cohort (16%). Genetic
influences on OCD have been studied mainly
using the family history method, sensitive
* Address for correspondence: Derek Bolton, Ph.D., Psychology
Department, Institute of Psychiatry PO Box 77, De Crespigny Park,
London SE5 8AF, UK.
Psychological Medicine, Page 1 of 10.
f 2006 Cambridge University Press
Printed in the United Kingdom
to familial aggregation, but not directly to the
distinction between genetic and shared environ-
mental influences. Family studies in the 1980s/
1990s consistently found that relatives of OCD
probands had raised rates of OCD or sub-
threshold OCD compared to non-clinic controls
ranging between 4% and 35%, this large range
partly reflecting variable criteria for ‘sub-
threshold’ OCD (Lenane et al. 1990; Riddle
et al. 1990; Bellodi et al. 1992; Leonard et al.
1992; Thomsen, 1995). More recent family his-
tory studies have suggested that familial aggre-
gation of OCD may apply only to early-onset
OCD (<18 years) (Nestadt et al. 2000; Carter
et al. 2004), and even then not in all cases
(Chabane et al. 2005).
Much of the family history research on OCD
has focused on co-morbidity and its impli-
cations for genetic association. Following the
finding that paediatric OCD is often associated
with tics, a seminal family history study by Pauls
and colleagues (1986) used a novel method de-
pending on proband co-morbidity. They found
that probands with tic disorder without OCD
were as likely to have relatives with OCD as
probands with tic disorder but with OCD, from
which it was inferred that OCD is an alternate
expression of the genes for tic disorder. A fur-
ther study by the Yale group (Pauls et al. 1995)
began with OCD probands and examined rates
of tics and OCD and sub-threshold OCD in
first-degree relatives compared with non-clinical
controls, finding higher rates of tics among
relatives of OCD probands than controls (4.6%
v. 1.0%). The study also found that relatives of
early-onset OCD cases (<18 years) were more
likely to have both OCD and tics, with greatest
risk in case of onset between ages 5 and 9 years.
This suggests that childhood onset OCD may be
the more heritable form (associated with tics).
Finally, the Yale group study found that rates of
tics were higher in relatives of OCD probands
with a family history of OCD compared with
those without family history, consistent with the
hypothesis that tic-related OCD is a distinctive
familial subtype. Subsequent family history
studies have supported the hypothesis that tic
disorders constitute an alternate expression
of the familial OCD phenotype (Grados et al.
2001; Hanna et al. 2005). Much of this work
has been with clinic samples, which may have
greater co-morbidity. Two studies of early-onset
OCD in community cohorts have failed to find
association between tics and the disorder
(Flament et al. 1988; Douglass et al. 1995).
Family history methodology has also been
used to examine the association between OCD
and other anxiety disorders. Family history
studies of OCD that included an anxious non-
OCD control group did not find raised rates of
obsessive–compulsive symptoms in first-degree
relatives compared with those controls (Clark &
Bolton, 1985; Last et al. 1991). A family study
by Black and colleagues (1992) found that the
rates of sub-threshold OCD and anxiety dis-
order were higher among the relatives of adult
probands with OCD compared with relatives
of psychiatrically normal controls, suggesting
that anxiety disorder diathesis is transmitted in
families with OCD, consistent with the classi-
fication of OCD as an anxiety disorder.
Subsequent analysis using the methodology of
Pauls et al. (1986) described above compared
psychiatric disorders in the relatives of pro-
bands with OCD and non-clinic controls, find-
ing a raised rate of generalized anxiety disorder
(GAD) in OCD relatives of OCD probands,
including those without co-morbid GAD, sug-
gesting that GAD may represent an alternative
expression of the genetic factor(s) contributing
to OCD (Black et al. 1995). Recent study using
similar family history methodology also with
adult samples have found familial aggregation
of OCD and GAD (Nestadt et al. 2001), and of
OCD and anxiety disorders, more strongly for
OCD with early onset (<10 years) (Carter et al.
2004). To date there have been no reported
family history studies of OCD co-morbidity
with anxiety disorders for child probands, and
no reports of OCD co-morbidity with anxiety
disorders or tics using twin methodology.
The present twin study was designed to
address the following sets of questions. First, is
there familial aggregation of early-onset OCD
using well-defined criteria for sub-threshold
caseness in a population sample, and if so, to
what extent is this attributable to genetic influ-
ence? Second, is early-onset OCD associated
with tics in a community sample, and is there
familial aggregation of these two conditions?
Is there also familial aggregation of OCD and
anxiety disorders? Third, to what extent is
familial aggregation of these three conditions
attributable to genetic influence? The study
2D. Bolton et al.
design and sample also allowed us to obtain an
estimate of heritability of tics in a community
Design: two-phase stratified sampling
The present study is part of the Institute of
Psychiatry’s Genetic and Environmental Effects
on Emotion Study (GEMS), and the basic
methodology and procedure have been reported
elsewhere (Bolton et al. 2006). In summary, a
standard two-phase stratified sampling design
was used, the first phase comprising 4662 twin
pairs in the Twins Early Development Study
(TEDS; Trouton et al. 2002). Of the 4662 twin
pairs in the Phase 1 sample, 754 were mono-
zygotic male (MZM), 783 dizygotic male
(DZM), 845 monozygotic female (MZF), 768
dizygotic female (DZF), and 1512 were dizy-
gotic opposite-sex (DZO) pairs. The Phase 1
sample were screened at age 4 using a maternal-
informant composite questionnaire on anxiety-
related behaviours and tics, and a high-risk
sample defined by scores on the screening
instrument were selected, along with a control
group, for detailed assessment in Phase 2 at age
6 years, in the first half of their seventh year.
The selection criterion identified 1833 screen-
positive twin pairs, that is, twin pairs with at
least one of the twins screen-positive, with the
remaining 2829 pairs being screen-negative, that
is, pairs with neither child screen-positive. Of
the 1833 screen-positive twin pairs 1296 were
available for the present study, and of the 2829
screen-negative pairs 2318 were available, the
others being unavailable because of partici-
pation in other studies. Thus, for assessment
in Phase 2, all available 1296 screen-positive
twin pairs were selected for study, together with
a control group of 192 screen-negative twin
pairs selected randomly from the 2318 available,
making a total of 1488 twin pairs (2976
In two-phase sampling designs Phase 2 data-
points are adjusted by selection probability
strata weights, computed as the inverse of the
proportion of the number of Phase 2 obser-
vations in a given stratum (defined by scores on
the Phase 1 screen) to the number in that
stratum in the Phase 1 sample. The standard
principle for epidemiological surveys by which
selection probability strata weights are attached
to the data-points of individuals (observed in
Phase 2) has to be qualified in the case of twin
population samples. Because in twin analyses
pairs and not individuals are the units of study,
weights are assigned to pairs. In the case of a
binary classification, as used in the present
study, the principle is that a twin pair is assigned
a screen-positive weight in case either co-twin,
but not necessarily both, is screen-positive, and
is assigned a screen-negative weight in case both
co-twins are screen-negative. The same selection
probability weights are used for estimating
Ethical approval for the study was given by
the Research Ethics Committee of the Institute
of Psychiatry and South London and Maudsley
NHS Trust, and written informed consent was
obtained from the mothers participating in the
Phase 2 observed sample
The sample observed in Phase 2 consisted of
1708 children, aged between 6 and 61
comprising 854 twin pairs, 253 MZ, and 601
DZ. Response rate was 57% (n=854/1488 twin
pairs). Main reasons for non-participation in
Phase 2 were, as percentages of the whole selec-
ted sample (n=1488 families) and in order of
size of the groups, as follows: failure to return
written consent forms or failure to arrange
the interview before the time window for the
assessment elapsed (i.e. before twins exceeded 61
years) (20%), untraceable (12%), ‘too busy to
participate’ (5%), with the remaining 5% due
to miscellaneous other reasons.
Assessments in Phase 2
In Phase 2 diagnostic status according to
DSM-IV was assessed by telephone interview
with the mothers using the parental version of
the Anxiety Disorders Interview Schedule for
Children and Parents (ADIS-C/P; Silverman
& Nelles, 1998), amended to assess lifetime
diagnoses. In the present study a distinction
was made between ‘symptom syndrome’ and
‘diagnosis’. A child was assigned a ‘symptom
syndrome’ where they met the full DSM-IV
symptom criteria for a disorder regardless of
degree of impairment, defined according to
ADIS-C/P rules. Diagnoses were only assigned
wherethechildmet boththe symptom
Obsessive compulsion in 6-year-old twins3
syndrome criterion and associated impairment
criterion. Full details of amendment and use of
the ADIS-C/P are given in our previous report
(Bolton et al. 2006).
A section on tics was appended to the diag-
nostic interview, obtained from the Yale Child
Study Center (J. Leckman, personal communi-
cation, 1999). This comprised a paragraph ex-
plaining to the interviewee what tics are, and a
subsequent explanatory paragraphs on motor
tics, followed by the question: ‘Has your child
ever had (or does he/she now have) facial tics,
jerks of other parts of the body, or any unusual
movements or habits?’ If the answer to this
question was ‘yes’, information was sought on
the nature of the tic, on frequency, including
whether nearly every day, onset, including
whether at least a month ago, or at least a year
ago, duration, including whether for at least 4
months or at least a year, and whether there has
ever been a tic-free period of at least three con-
secutive months. There follows a subsection on
vocal tics, prefaced by an explanatory para-
graph, then the question: ‘Have you ever found
(or do you find now) your child making invol-
untary noises other than normal talking, like
grunts, throat clearing, or saying words or part
of words?’ If the answer to this question was
‘yes’, information was sought as above on the
nature, frequency, and duration of tics and of
tic-free periods. This information is sufficient to
establish according to DSM-IV criteria presence
of absence of transient tic disorder, chronic
motor or vocal tic disorder and Tourette’s dis-
Phenotypes examined in the present study
Symptom syndrome is a plausible phenotype in
genetic studies, and has the advantage that it is
more common than diagnosis, providing greater
statistical power to detect effects. To maximize
numbers, lifetime symptom syndrome pheno-
types were selected for consideration in the case
of the rarer conditions, OCD and tics. In the
case of OCD, again to maximize numbers, a
phenotype sub-threshold for symptom syn-
drome (and diagnosis) was also considered, de-
fined straightforwardly using ADIS/C-P rules,
as follows: (a) positive response to screen for
obsessions or for compulsions, and (b) positive
response to ‘persistence’; but without requiring
significant distress at the symptom level or re-
sistance (trying to stop). In the case of tics, we
selected as the tic disorder phenotype transient
plus chronic tic disorder lifetime symptom syn-
drome (that is, requiring motor or phonic tics
of at least 4 weeks duration). A general category
of any anxiety disorder other than OCD was
defined, comprising separation anxiety disorder,
social phobia, specific phobia, and GAD, post-
traumatic stress disorder and panic disorder.
This general category was much more common
in the sample than tics or OCD, and to reduce
this effect any anxiety disorder lifetime diag-
nosis (rather than syndrome) was selected for
consideration. Thus the following three con-
ditions were considered in testing hypotheses in
the present study: OCD lifetime sub-threshold
syndrome, tic disorder lifetime syndrome, and
any anxiety disorder (other than OCD) lifetime
diagnosis. Kappa coefficients for assessment
of these three conditions were obtained using
the method described in our previous report
(Bolton et al. 2006) and were as follows: 0.80 for
OCD lifetime sub-threshold syndrome, 0.80 tic
disorder lifetime syndrome, and 0.88 for any
anxiety disorder (other than OCD) lifetime
For these three conditions weighted preva-
lence rates in the whole sample (n=9324
individuals), with 95% confidence intervals
(CIs), and with raw (unweighted) numbers in
the Phase 2 sample (n=1708 individuals) in
square brackets, were as follows: for OCD life-
time sub-threshold syndrome (6.1%, 95% CI
4.2–8.5 ); for tic disorder lifetime syndrome
(6.0%, 95% CI 4.2–8.4 ); and for any
anxiety disorder (other than OCD) lifetime
diagnosis (19.1%, 95% CI 16.1–22.4 ). For
comparison purposes, the rate for OCD lifetime
syndrome was 2.5% (95% CI 1.4–4.3 ),
indicating that the broadening of the phenotype
to include sub-syndromal cases approximately
doubled the prevalence rate.
All analyses incorporated selection probability
weights as defined above to account for selec-
tion of the sample assessed in Phase 2. Preva-
lence rates were computed on weighted data
using STATA statistical software for survey data
(Stata Corporation, 2004), which also permitted
control for the effects of ‘clustering’ of con-
ditions in the twin pairs. To investigate the
4 D. Bolton et al.
hypotheses of the study, multivariate genetic
analyses were applied to three variables: OCD
lifetime sub-threshold syndrome, tic disorder
lifetime syndrome, and any anxiety disorder
(other than OCD) lifetime diagnosis. These
conditions are defined above and for brevity
are shortened, using italics, to OCD, Tics, and
Anxiety disorder. Analytical methodology is as
Liability-threshold model fitting
Since all variables were dichotomous (‘no’=0,
‘yes’=1), tetrachoric correlations and par-
ameters of the genetic models were estimated
using liability-threshold models fitted in the
programme Mx (Neale, 1999). The assumption
is that each symptom dimension has an under-
lying normal distribution of liability with a dis-
crete threshold (above which a child is assumed
to show the condition of relevance) and that the
joint liabilities (e.g. twin 1 and twin 2 scores)
follow a bivariate normal distribution where
both traits have a mean of 0 and standard devi-
ation 1. The correlation between the liabilities
and the thresholds are estimated from the rela-
tive proportions in each category (i.e. number
of pairs where both twins score ‘yes’, or ‘no’, or
are discordant). For the simultaneous analyses
of three or more dichotomous variables Mx
estimation. Essentially, the model predicts pro-
portions of twin pairs that should exist for the
various possible patterns of responses from two
twins, assuming a multivariate normal distri-
bution. These expected proportions are then
evaluated against the observed cell proportions
to derive a maximum-likelihood correlation
matrix for both MZ and DZ twin pairs.
Tetrachoric correlations between OCD, Tics and
The MZ and DZ correlations between the six
dichotomous traits (i.e. OCD, Tics and Anxiety
disorder for twin 1 and twin 2) were estimated
in a constrained model that produced a reduced
number of correlations to simplify interpret-
ation: three within-twin cross-trait correlations
(equal across MZ and DZ pairs); three cross-
twin within-trait correlations (MZ and DZ
pairs, separately); and three cross-twin cross-
trait correlations (MZ and DZ pairs, separ-
ately). These correlations will indicate if there
are common aetiological influences on the three
liabilities and to what extent these influences
are due to overlapping genetic, shared-environ-
mental and non-shared-environmental effects.
These sources of covariance are formally tested
in the genetic model (see below).
Genetic model fitting
Genetic model fitting of twin data estimates
the contribution of genetic and environmental
influences to individual differences in a trait.
Three latent components are inferred from the
data: additive genetic influence (A), shared
or common environment (C), and non-shared
environment (E). Identical (MZ) share all their
genes whilst non-identical (DZ) twins share only
50% of their genes. Assuming that MZ and DZ
twins are equally similar in terms of their
environment, then any excess of similarity be-
tween MZ twins over DZ twins is assumed to
be due to the greater genetic sharing for MZ
twins, thus giving estimates of genetic effects.
Resemblance between MZ twins not due to
genetic effects is assumed to be due to the shared
environment (C), and differences between MZ
twins are attributable to differing environmental
impacts on the individuals and thus permit
estimate of E. The relative magnitude and im-
portance of these latent factors can be inferred
by fitting the raw ordinal data (with observed
correlational patterns) to the predicted corre-
lations according to the hypothesized model
(ACE, AE, CE or E). Full details of this method
are given elsewhere (Neale & Cardon, 1992;
Plomin et al. 2001).
In addition, when multiple liabilities are
measured, their associations (implying common
aetiological influences) can be partitioned into
genetic, shared-environmental and non-shared-
environmental correlations. The power to dis-
tinguish between these different sources of
co-occurrence is derived from the MZ/DZ ratio
of these correlations: a 2:1 ratio is indicative
of additive genetic effects, whereas a 1:1 ratio
suggests influences of common environment in
inducing a correlation between two variables.
Non-significant cross-trait cross-member corre-
lations imply that the common aetiological
influences are due to individual specific en-
vironment (E), not familial effects.
A full Cholesky ACE decomposition was
fitted to the data, and nested sub-models were
Obsessive compulsion in 6-year-old twins5
evaluated. Since analyses on raw data produces
a log-likelihood of the data, to obtain an overall
measure of fit (x2), we need to compute the
difference in likelihoods between each genetic
model and a perfect fitting (saturated) model
in which the maximum number of parameters is
estimated to describe the correlational structure
between all variables. We have only reported the
x2indices here (Table 2). The fit of all models
were assessed by the x2statistic and degrees
of freedom of the model; a non-significant x2
indicates a good-fitting model. The fit of nested
sub-models (e.g. AE or CE) compared with the
full model were evaluated by changes in x2
relative to the associated change in degrees of
freedom, and comparison of sub-models with
the same number of parameters used Akaike’s
Information Criterion (AIC, x2x2 df), by which
higher negative values indicate a better fit
(Neale & Cardon, 1992). Information about the
precision of parameter estimates was obtained
by likelihood-based CIs (Neale & Miller, 1997).
The first set of questions the present study
aimed to address is whether individual differ-
ences in paediatric OCD, tics and anxiety
(due to either genetic or shared environmental
influences). Relevant results here, as for other
questions addressed, are all from the multi-
variate modelling. Table 1 shows tetrachoric
correlations within and cross twins, and within
Table 1. Tetrachoric correlations within- and cross-twins, and within- and cross-traits
(with 95% CI) for OCD, Tics and Anxiety disorder
Within-twin cross-trait Cross-twin within-traitCross-twin cross-trait
Whole sampleMZ pairs DZ pairsWhole sampleMZ pairs DZ pairs
The within-twin cross-trait correlations were constrained cross twins such that, e.g. OCDtwin1xTicstwin1=OCDtwin2xTicstwin2. The cross-
twin cross-trait correlations are constrained such that, e.g. OCDtwin1xTicstwin2=OCDtwin2xTicstwin1. Confidence intervals including zero
indicate non-significance. Thresholds for MZ pairs were: 1.50, 1.47, 0.64, and for DZ pairs: 1.37, 1.47, 0.64.
Table 2.Multivariate genetic model-fitting results (raw ordinal analyses with weights)
(5) No rg+rcOCD–Tics
(6) No rg+rcOCD–Anxiety disorder
(7) No rg+rcTics–Anxiety disorder
Models 5, 6 and 7 test the significance of a familial correlation between the traits. rgand rcare the genetic and shared-environmental
correlation, respectively and are reported in Table 4. Dx2is derived by comparing sub-models to the full ACE model. The critical x2value (at
the 0.05 level) for 1 df is 3.84.
AIC, Akaike’s Information Criterion (x2x2 df), so that higher negative values indicate a better fit.
* Significant decline in fit.
6D. Bolton et al.
and cross traits, with 95% CIs, for the three
phenotypes OCD, Tics and Anxiety disorder.
The higher MZ compared to DZ cross-twin
within-trait correlations (second column in
Table 1), indicate genetic influences, especially
for OCD and Tics. Results of the formal testing
of the significance of genetic and environmental
effects are presented in Table 2.
Dropping either all genetic parameters (for
OCD, Tics and Anxiety disorder) or all shared-
environmental parameters (models 2 and 3) did
not result in a significant decline in fit [Dx2
(6 df)=5.35, p=0.49, and 5.8, p=0.45 respect-
ively]. However, dropping A and C at the same
time showed a highly significant deterioration in
fit [Dx2(12 df)=477, p<0.001]. This means that,
although there are highly significant familial
effects on each of the traits, our sample, in terms
of size, lacks the power to detect these influences
separately. Standardized estimates of the genetic
and environmental variance of OCD, Tics and
Anxiety disorder (based on the full ACE model)
are presented on the diagonals of Table 3.
Familial effects (combining genetic and shared
environmental effects) range from 47% (OCD)
to 61% (Tics).
The second set of questions to be examined in
the present study concerns associations between
OCD and tics, and OCD and anxiety disorders
in paediatric community samples. It can be seen
in the bottom half of Table 1, first column,
that the correlation between the phenotypes
OCD and Tics is 0.32, and that this difference is
statistically significant since the lower bound
of confidence interval, which is based on x2dif-
ference tests at the 0.05 level, exceeds 0. The
correlation between the phenotypes OCD and
Anxiety disorder is 0.25, and this is also statisti-
cally significant at the 0.05 level. The correlation
between the phenotypes Tics and Anxiety dis-
order is lower, 0.14, although still statistically
significant. Evidence of familial aggregation
of the three disorders can be derived from the
presence of significant cross-twin cross-trait
correlations. The fourth column in Table 1
shows that all these correlations are significant,
with the largest aggregations for OCD and Tics
and OCD and Anxiety disorder. Phenotypic
associations may also be expressed in terms of
odds ratios (ORs), as follows, with 95% CIs and
significance values: OCD/Tics (OR 4.4, 95%
CI 1.3–15.2, p=0.018), OCD/Anxiety disorder
Standardized estimates (with 95% CI) based on the full ACE model
On the diagonals of the A column, are the heritabilities of the traits and on the diagonals of the C and E columns the standardized shared-environmental variance (c2), and non-shared
environmental variance (e2), respectively. On the off-diagonals are the phenotypic correlations due to A, C, E (with 95% CI), with the proportions given in square brackets. These values are a
function of both the standardized estimates on the diagonals and the A, C and E correlations given in Table 4, and add up to the phenotypic correlation (last column Table 4).
Obsessive compulsion in 6-year-old twins7
(OR 2.9, 95% CI 1.0–8.4, p=0.046), and
Anxiety disorder/Tics (OR 2.5, 95% CI 1.1–5.7,
p=0.026). (The contingency tables on which
these odds ratios are based are available from
the corresponding author.)
The third set of questions to be considered is
to what extent familial aggregations of these
three conditions are attributable to genetic or
shared environmental influences. For this, we
look first at the third column of Table 1. Higher
MZ compared to DZ cross-twin cross-trait
correlations indicate a significant proportion of
genetic influences to explain the co-occurrence
of the conditions. However, the table shows
MZ/DZ cross-twin cross-trait correlations that
are quite similar for MZ and DZ pairs, in-
dicating that shared-environmental factors pre-
dominantly determine the familial aggregation
The first three columns of Table 4 show the A,
C and E correlations across disorders (rg, rc, re),
i.e. the extent to which the same A, C and E
factors influence the disorders. The off-diagonal
elements are a function of rg, rc, re and the
standardized variance components given on the
diagonals of Table 3. It can be seen in Table 4
that the genetic (rg), shared-environmental (rc)
and non-shared-environmental (rc) correlations
for all combinations of symptoms have a
zero lower 95% CI, indicating non-significance.
This is also the case for the proportions of the
phenotypic correlations (rph) explained by A, C
and E (given on the off-diagonals of Table 3).
However, formal testing of the combined effects
of genetic and shared-environmental corre-
lations (Table 2, models 5–7) showed significant
familial aggregation of OCD and Tics and OCD
and Anxiety disorder [Dx2(2 df)=10.4, p=0.005
and 15.2, p=0.001 respectively), though not of
Tics and Anxiety disorder. This means that
although there is significant familial aggregation
between OCD and Tics, and between OCD and
Anxiety disorder (80% and 97% respectively),
we lacked the power to determine the specific
Previous estimates of heritability of OCD
assessed with symptom scales have suggested
modest to moderate additive genetic effects, in a
range between 26% and 58%, as reviewed in the
Introduction. The present estimate of additive
genetic effects of 29% on an OCD phenotype
defined in diagnostic terms, but including sub-
threshold cases, is consistent with the lower end
of this range, and the estimate of familial ag-
gregation due to combined additive genetic and
shared environment effects, which could not be
distinguished in this study, as 47%, is consistent
with the upper end of the range of previous
Evidence for genetic factors in tic disorders
has come from reports of high concordance in
MZ twin pairs and lower concordance in DZ
twins (e.g. Price et al. 1985; Hyde et al. 1992).
The present large-scale twin study in a com-
munity sample estimates heritability as 50%
for tic disorder syndrome in this population.
This estimate is based on a full ACE model in
which neither genetic nor shared environmental
effects are statistically significant, and the com-
bined effect of both sources of familial aggre-
gation account for 61% of individual variation.
Evidence for the validity of the assessment of
the phenotype in the present study is that the
prevalence estimate for Tics of 6.0% (95% CI
4.2–8.4%) is broadly consistent with previous
epidemiological studies of tics; for recent review
see Scahill et al. (2001).
The heritability estimate for the complex
phenotype ‘any anxiety disorder’ is low and not
statistically significant. Our recent report on
the same sample found high and statistically
Table 4. Genetic (rg), shared-environmental (rc) non-shared environmental (re) and
phenotypic correlation (rph) (with 95% CI) based on the full ACE model
Confidence intervals including zero indicate non-significance.
8 D. Bolton et al.
significant additive genetic effects for specific
phobia and separation anxiety disorder (Bolton
et al. 2006). Notwithstanding the fact that in this
sample, these two disorders constituted most
of the complex category ‘any anxiety disorder’
inclusion of the rarer anxiety disorders had
the effect of substantially reducing familial
aggregation and estimate of heritability.
There were significant within-twin associ-
ations between OCD and tics, this being the first
community population study to replicate this
pattern previously found in clinic samples.
Significant within-twin associations were also
found between OCD and anxiety disorder in
this paediatric population, and to a lesser extent
between tics and anxiety disorder These pheno-
typic associations were also found across-twins
indexing familial aggregation for OCD and Tics
and for OCD and Anxiety disorder, but not
for Tics and Anxiety disorder. This pattern of
findings is consistent with the hypothesis of a
tic-related early-onset OCD phenotype, but
also with the hypothesis of an anxiety-related
early-onset OCD phenotype, with, however,
some overlap between them.
Limitations of the study include that diag-
nostic interview assessment relied on maternal-
behavioural observation data, which may have
led to under-reporting particularly of obsessions
and tics. There are also limitations inherent in
the twin design and the ACE models for the
purpose of heritability estimates. These include
chorionicity, atypical gestation of MZ twins,
and increased similarity of environment for MZ
twins as compared to DZ twins (Martin et al.
1997), as well as the inclusion of gene–environ-
ment correlations and interactions in the genetic
parameter. These limitations are varied in
their effects, some resulting in conservative
heritability estimates, others resulting in inflated
heritability estimates. Finally, study of diag-
nostically defined phenotypes has the disadvan-
tage that modelling of dichotomous traits has
less power than modelling of continuous traits
and is less able to distinguish between genetic
and shared environmental factors (Neale et al.
1994). This limitation affected the modelling
in the present study, and it has to be weighed
against the advantage of diagnostically defined
phenotypes of validity in relation to psycho-
This study was supported by The Wellcome
Trust, grant 056163.
DECLARATION OF INTEREST
Bellodi, L., Sciuto, G., Diaferia, G., Ronchi, P. & Smeraldi, E. (1992).
Psychiatric disorders in the families of patients with obsessive-
compulsive disorder. Psychiatry Research 42, 111–120.
Black, D. W., Goldstein, R. B., Noyes, R. & Blum, N. (1995).
Psychiatric disorders in relatives of probands with obsessive-
compulsive disorder and co-morbid major depression or general-
ized anxiety. Psychiatric Genetics 5, 37–41.
Black, D. W., Noyes Jr., R., Goldstein, R. B. & Blum, N. (1992). A
family study of obsessive-compulsive disorder. Archives of General
Psychiatry 49, 362–368.
Bolton, D., Eley, T., O’Connor, T., Perrin, S., Rabe-Hesketh, S.,
Rijsdijk, F. & Smith, P. (2006). Prevalence and genetic and
environmental influences on anxiety disorders in 6-year-old twins.
Psychological Medicine 36, 335–344.
Carter, A. S., Pollock, R. A., Suvak, M. K., Pauls, D. L., Carter,
A. S., Pollock, R. A., Suvak, M. K. & Pauls, D. L. (2004). Anxiety
and major depression co-morbidity in a family study of obsessive-
compulsive disorder. Depression & Anxiety 20, 165–174.
Chabane, N., Delorme, R., Millet, B., Mouren, M. C., Leboyer, M. &
Pauls, D. (2005). Early-onset obsessive-compulsive disorder: a
subgroup with a specific clinical and familial pattern? Journal of
Child Psychology & Psychiatry 46, 881–887.
Clark, D. & Bolton, D. (1985). Obsessive-compulsive adolescents and
their parents: a psychometric study. Journal of Child Psychology
and Psychiatry 26, 267–276.
Clifford, C. A., Murray, R. M. & Fulker, D. W. (1984). Genetic and
environmental influences on obsessional traits and symptoms.
Psychological Medicine 14, 791–800.
Douglass, H. M., Moffitt, T. E., Dar, R., Mcgee, R. & Silva, P.
(1995). Obsessive-compulsive disorder in a birth cohort of 18-year-
olds: prevalence and predictors. Journal of the American Academy
of Child and Adolescent Psychiatry 34, 1424–1431.
Flament, M. F., Whitaker, A., Rapoport, J. L., Davies, M., Berg,
C. Z., Kalikow, K., Sceery, W. & Shaffer, D. (1988). Obsessive
compulsive disorder in adolescence: an epidemiological study.
Journal of the American Academy of Child and Adolescent
Psychiatry 27, 764–771.
Grados, M. A., Riddle, M. A., Samuels, J. F., Liang, K. Y., Hoehn-
Saric, R., Bienvenu, O. J., Walkup, J. T., Song, D. & Nestadt, G.
(2001). The familial phenotype of obsessive-compulsive disorder in
relation to tic disorders: the Hopkins OCD family study.
Biological Psychiatry 50, 559–565.
Hanna, G. L., Himle, J. A., Curtis, G. C. & Gillespie, B. W. (2005). A
family study of obsessive-compulsive disorder with pediatric pro-
bands, American Journal of Medical Genetics. B: Neuropsychiatric
Genetics 134B, 13–19.
Hettema, J. M., Neale, M. C. & Kendler, K. S. (2001). A review and
meta-analysis of the genetic epidemiology of anxiety disorders.
American Journal of Psychiatry 158, 1568–1578.
Hudziak, J. J., Van Beijsterveldt, C. E., Althoff, R. R., Stanger, C.,
Rettew, D. C., Nelson, E. C., Todd, R. D., Bartels, M. & Boomsma,
D. I. (2004). Genetic and environmental contributions to the Child
Behavior Checklist Obsessive-Compulsive Scale: a cross-cultural
twin study, Archives of General Psychiatry 61, 608–616.
Hyde, T. M., Aaronson, B. A., Randolph, C., Rickler, K. C. &
Weinberger, D. R. (1992). Relationship of birth weight to the
phenotypic of Gilles de la Tourette’s Syndrome in monozygotic
twins. Neurology 42, 652–658.
Obsessive compulsion in 6-year-old twins9
Jonnal, A. H., Gardner, C. O., Prescott, C. A. & Kendler, K. S. Download full-text
(2000). Obsessive and compulsive symptoms in a general popu-
lation sample of female twins. American Journal of Medical
Genetics 96, 791–796.
Last, C. G., Hersen, M., Kazdin, A. E., Orvaschel, H. & Perrin, S.
(1991). Anxiety disorders in children and their families. Archives of
General Psychiatry 48, 928–934.
Lenane, M. C., Swedo, S. E., Leonard, H., Pauls, D. L., Sceery, W. &
Rapoport, J. L. (1990). Psychiatric disorders in first degree
relatives of children and adolescents with obsessive compulsive
disorder. Journal of American Academy of Child Adolescent
Psychiatry 29, 407–412.
Leonard, H. L., Lenane, M. C., Swedo, S. E., Rettew, D. C., Gershon,
E. S. & Rapoport, J. L. (1992). Tics and Tourette’s disorder: a 2 to
7 follow-up of 54 obsessive-compulsive children. American Journal
of Psychiatry 149, 1244–1251.
Martin, N., Boosma, D. & Machin, G. (1997). A twin-pronged attack
on complex traits, Nature Genetics, 17, 387–392.
Neale, M. C. (1999). Mx: Statistical Modeling (5th edn). Box 126
MCV, Richmond, VA 23298: Department of Psychiatry.
Neale, M. C. & Cardon, L. (1992). Methodology for Genetic
Studies of Twins and Families. Kluwer Academic Publications:
Neale, M., Eaves, L. J. & Kendler, K. S. (1994). The power of the
classical twin study to resolve variation in threshold traits.
Behavior Genetics 24, 239–258.
Neale, M. C. & Miller, M. B. (1997). The use of likelihood based
confidence intervals in genetic models. Behavior Genetics 27,
Nestadt, G., Samuels, J., Riddle, M., Bienvenu III, O. J., Liang, K. Y.,
LaBuda, M., Walkup, J., Grados, M. & Hoehn-Saric, R. (2000).
A family study of obsessive-compulsive disorder, Archives of
General Psychiatry 57, 358–363.
Nestadt, G., Samuels, J., Riddle, M., Liang, K. Y., Bienvenu, O. J.,
Hoehn-Saric, R., Grados, M. & Cullen, B. (2001). The relationship
between obsessive-compulsive disorder and anxiety and affective
disorders. Psychological Medicine 31, 481–487.
Pauls, D., Towbin, K., Leckman, J., Zahner, G. & Cohen, D.
(1986). Gilles de la Tourette syndrome and obsessive-compulsive
disorder: evidence supporting an etiological relationship. Archives
of General Psychiatry 43, 1180–1182.
Pauls, D. L., Alsobrook, J. P., Goodman, W., Rasmussen, S. &
Leckman, J. F. (1995). A family study of obsessive-compulsive
disorder. American Journal of Psychiatry 152, 76–84.
Plomin, R., DeFries, J. C., McClearn, G. & McGuffin, P. (2001).
Behavioral Genetics (4th edn). Worth Publishers: New York.
Price, R. A., Kidd, K. K., Cohen, D. J., Pauls, D. L. & Leckman, J. F.
(1985). A twin study of Tourette Syndrome. Archives of General
Psychiatry 42, 815–820.
Riddle, M. A., Scahill, L., King, R., Hardin, M. T., Towbin, K. E.,
Ort, S. I., Leckman, J. F. & Cohen, D. J. (1990). Obsessive
compulsive disorder in children and adolescents: phenomenology
and family history. Journal of American Academy of Child
Adolescent Psychiatry 29, 766–772.
Scahill, L., Tanner, C. & Dure, L. (2001). The epidemiology of tics
and Tourette Syndrome in children and adolescents. In Advances
in Neurology, vol. 85 (ed. D. J. Cohen, J. Jankovic, C. G. Goetz),
pp. 261–271. Lippincott Williams and Wilkins. Philadelphia, PA.
Shih, R. A., Belmonte, P. L. & Zandi, P. P. (2004). A review of the
evidence from family, twin and adoption studies for a genetic
contribution to adult psychiatric disorders. International Review of
Psychiatry 16, 260–283.
Silverman, W. K. & Nelles, W. B. (1998). The anxiety disorders
interview schedule for children. Journal of American Academy of
Child and Adolescent Psychiatry 27, 772–778.
Stata Corporation (2004). STATA version 8. Stata Corporation, TX,
Thomsen, P. H. (1995). Obsessive-compulsive disorder in children
and adolescents: a study of parental psychopathology and pre-
cipitating events in 20 consecutive Danish cases. Psychopathology
Trouton, A., Spinath, F. M. & Plomin, R. (2002). Twins early
development study (TEDS): a multivariate, longitudinal genetic
investigation of language, cognition and behavior problems in
childhood. Twin Research 5, 444–448.
van Grootheest, D. S., Cath, D. C., Beekman, A. T. & Boomsma, D. I.
(2005). Twin studies on obsessive-compulsive disorder: a review.
Twin Research and Human Genetics 8, 450–458.
10D. Bolton et al.