Longitudinal stability of the CBCL-juvenile bipolar disorder phenotype: A study in Dutch twins.
ABSTRACT The Child Behavior Checklist-juvenile bipolar disorder phenotype (CBCL-JBD) is a quantitative phenotype that is based on parental ratings of the behavior of the child. The phenotype is predictive of DSM-IV characterizations of BD and has been shown to be sensitive and specific. Its genetic architecture differs from that for inattentive, aggressive, or anxious-depressed syndromes. The purpose of this study is to assess the developmental stability of the CBCL-JBD phenotype across ages 7, 10, and 12 years in a large population-based twin sample and to examine its genetic architecture.
Longitudinal data on Dutch mono- and dizygotic twin pairs (N = 8013 pairs) are analyzed to decompose the stability of the CBCL-JBD phenotype into genetic and environmental contributions.
Heritability of the CBCL-JBD increases with age (from 63% to 75%), whereas the effects of shared environment decrease (from 20% to 8%). The stability of the CBCL-JBD phenotype is high, with correlations between .66 and .77 across ages 7, 10, and 12 years. Genetic factors account for the majority of the stability of this phenotype. There were no sex differences in genetic architecture.
Roughly 80% of the stability in childhood CBCL-JBD is a result of additive genetic effects.
- [Show abstract] [Hide abstract]
ABSTRACT: Background: The clinical presentation of children and adolescents referred to mental health services is frequently complicated by comorbid and severe affective and behavioral dysregulation. This dysregulation phenotype seems to be an indicator of overall psychopathology, symptom severity and functional impairment. Currently, this phenotype is assessed by the Child Behavior Checklist. However, the widely used Strengths and Difficulties Questionnaire (SDQ) has been recently validated to screen the Dysregulation Profile (SDQ-DP) in clinical settings. The objective of this study was to determine the prevalence and demographic, psychosocial and clinical correlates of the SDQ-DP phenotype in a Spanish clinical sample. Sampling and Methods: In a clinical sample of 623 consecutively referred children and adolescents (4-17 years old), we compared clinical and sociodemographic correlates between subjects who met the SDQ-DP criteria (DP) and those who did not (NO_DP). Sociodemographic data, parent-rated SDQ, Children's Global Assessment Scale, Clinical Global Impression, family Apgar scale and clinical diagnoses were collected by experienced child and adolescent psychiatrists. Results: Overall in our sample, 175 subjects (28.1%) met the SDQ-DP criteria (DP group). Compared with the NO_DP group, the DP subjects had significantly higher scores on internalizing and externalizing psychopathology, problems with peers and overall problems as well as significantly lower scores on prosocial behavior. Clinical diagnoses assigned revealed that DP subjects showed significantly greater psychiatric comorbidity. DP subjects also showed significantly worse family functioning and increased symptom severity and significantly lower scores on psychosocial functioning. Conclusions: A high prevalence of children and adolescents with the dysregulated profile, assessed by the SDQ-DP, was found in our clinical setting. The SDQ-DP may serve as an index of overall psychological severity and functional impairment. In addition, it may indicate family dysfunction. Further research is needed to validate the clinical value of SDQ-DP by examining longitudinal stability, heritability, adult outcome, risk factors and diagnostic correlates. © 2014 S. Karger AG, Basel.Psychopathology 05/2014; · 1.56 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Children meeting the Child Behavior Checklist Dysregulation Profile (CBCL-DP) suffer from high levels of co-occurring internalizing and externalizing problems. Little is known about the cognitive abilities of these children with CBCL-DP. We examined the relationship between CBCL-DP and nonverbal intelligence. Parents of 6,131 children from a population-based birth cohort, aged 5 through 7 years, reported problem behavior on the CBCL/1.5-5. The CBCL-DP was derived using latent profile analysis on the CBCL/1.5-5 syndrome scales. Nonverbal intelligence was assessed using the Snijders Oomen Nonverbal Intelligence Test 2.5-7-Revised. We examined the relationship between CBCL-DP and nonverbal intelligence using linear regression. Analyses were adjusted for parental intelligence, parental psychiatric symptoms, socio-economic status, and perinatal factors. In a subsample with diagnostic interview data, we tested if the results were independent of the presence of attention deficit hyperactivity disorder (ADHD) or autism spectrum disorders (ASD). The results showed that children meeting the CBCL-DP (n = 110, 1.8 %) had a 11.0 point lower nonverbal intelligence level than children without problems and 7.2-7.3 points lower nonverbal intelligence level than children meeting other profiles of problem behavior (all p values <0.001). After adjustment for covariates, children with CBCL-DP scored 8.3 points lower than children without problems (p < 0.001). The presence of ADHD or ASD did not account for the lower nonverbal intelligence in children with CBCL-DP. In conclusion, we found that children with CBCL-DP have a considerable lower nonverbal intelligence score. The CBCL-DP and nonverbal intelligence may share a common neurodevelopmental etiology.European Child & Adolescent Psychiatry 05/2014; · 3.70 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: El siguiente artículo revisa los principales instrumentos psicométricos utilizados para la evaluación de sintomatología asociada al trastorno bipolar infantil (TBPI). Se efectúa una revisión de la metodología asociada a la construcción del Kiddie Schedule for Affective Disorders and Schizophrenia for School - Age Children (KSADS), la Escala de Manía de Young (Young Mania Rating Scale [YMRS]) y del Child Mania Rating Scale (CMRS), analizando los indicadores y procedimientos estadísticos. Destacan las ventajas psicométricas del YMRS en el ámbito clínico y del CRMS en la investigación en torno a explorar el diagnóstico del trastorno bipolar en niños y adolescentes. Se discute la importancia de desarrollar investigación y trabajos que aborden el desarrollo de instrumentos para detectar sintomatología y epidemiología asociada a trastornos psiquiátricos, sobre todo para el caso del TBPI, cuya prevalencia en la población nacional es desconocida.Revista chilena de neuro-psiquiatría. 09/2013; 51(3):211-220.
Longitudinal Stability of the CBCL-Juvenile Bipolar
Disorder Phenotype: A Study in Dutch Twins
Dorret I. Boomsma, Irene Rebollo, Eske M. Derks, Toos C.E.M. van Beijsterveldt, Robert R. Althoff,
David C. Rettew, and James J. Hudziak
Background: The Child Behavior Checklist–juvenile bipolar disorder phenotype (CBCL-JBD) is a quantitative phenotype that is based
on parental ratings of the behavior of the child. The phenotype is predictive of DSM-IV characterizations of BD and has been shown
to be sensitive and specific. Its genetic architecture differs from that for inattentive, aggressive, or anxious–depressed syndromes. The
purpose of this study is to assess the developmental stability of the CBCL-JBD phenotype across ages 7, 10, and 12 years in a large
population-based twin sample and to examine its genetic architecture.
Methods: Longitudinal data on Dutch mono- and dizygotic twin pairs (N ? 8013 pairs) are analyzed to decompose the stability of
the CBCL-JBD phenotype into genetic and environmental contributions.
Results: Heritability of the CBCL-JBD increases with age (from 63% to 75%), whereas the effects of shared environment decrease (from
20% to 8%). The stability of the CBCL-JBD phenotype is high, with correlations between .66 and .77 across ages 7, 10, and 12 years.
Genetic factors account for the majority of the stability of this phenotype. There were no sex differences in genetic architecture.
Conclusions: Roughly 80% of the stability in childhood CBCL-JBD is a result of additive genetic effects.
Key Words: Childhood bipolar affective disorder, genetics, twins
the illness increasingly are being recognized as valid diagnoses,
how best to characterize children continues to be a focus of
extensive investigation (Biederman et al 1998; Faedda et al 1995;
Geller and Luby 1997; Weller et al 1995). One area of discussion
surrounds the degree to which the diagnosis of JBD in children
is associated with a more classic DSM-IV profile in adulthood
(Carlson et al 2000; Geller and Luby 1997; Leibenluft et al 2003).
For example, the adult-onset form of BD is associated with
discrete episodes of mania or hypomania and depression, whereas
this is not reported as common among children with JBD, in whom
the episodes are more often of long duration, with rapid cycling and
mixed mania (National Institute of Mental Health 2001). Given the
lack of prospective data (Faedda et al 2004), it is not surprising that
little is known about the developmental stability and change of JBD
using standard DSM approaches.
Several groups have described a JBD profile on the Child
Behavior Checklist (CBCL; Achenbach 1991) that differs from the
CBCL profiles of children with other DSM disorders (Biederman
et al 1995; Geller et al 1998; Wals et al 2001). Children with JBD
have a CBCL profile that includes high levels on the Aggressive
Behavior (AGG), Anxious/Depressed Behavior (A/D), and Atten-
tion Problems (AP) syndrome scales. The extent to which the
CBCL-JBD phenotype predicts DSM-III-R and DSM-IV diagnoses
of BD and delineates JBD from other childhood psychiatric
diagnoses such as attention-deficit/hyperactivity disorder (ADHD;
Biederman et al 1995), depression, and other disruptive behavior
he existence, prevalence, and taxonomy of juvenile bipo-
lar disorder (JBD) have been the focus of considerable
debate. Although the pediatric and adolescent forms of
disorders (Kahana et al 2003) has been examined across samples,
countries, and across methodologies (Althoff et al 2005). Results
have been replicated across age groups (Biederman et al 1995;
Carlson and Kelly 1998; Dienes et al 2002; Geller et al 1998;
Hazell et al 1999; Wals et al 2001), across treatment settings
(inpatient, outpatient), and across cultures (American, Dutch,
Brazilian, Australian). In a recent study, Faraone and colleagues
(2005) used a receiver-operating characteristic curve (ROC)
analysis on the profile of elevated AP, AGG, and A/D to predict
as in their siblings. The area under the curve statistic for prediction
of DSM-III-R BD using a summed score of the three scales was as
high as .97 for children with a current diagnosis of BD.
Recently, we described the prevalence of CBCL-JBD in a
general-population twin sample and contrasted the genetic ar-
chitecture of this phenotype (Hudziak et al, in press) to that of
AP. These data indicated that the prevalence of CBCL-JBD in
children is ?1% in boys and girls at ages 7, 10, and 12 years. At
each age, variation in CBCL-JBD was influenced by additive
genetic, shared, and unique environmental factors, with additive
genetic influences accounting for the largest part of the variance.
This is in contrast to the modeling of AP, which was influenced
by additive and nonadditive (dominance) genetic effects without
shared environmental influences. Liability threshold models of
CBCL-JBD versus CBCL-AP also showed that the CBCL-JBD
phenotype is unlikely to be an extreme version of CBCL-AP.
Latent class analyses of AGG, A/D, and AP symptoms in these
twins showed that a seven-class model fit best for girls, and an
eight-class model, best for boys. The most common class for both
boys and girls was one without symptoms. The CBCL-JBD severe
latent class was the least common—and was the only one that
had significant elevations on the suicidal items of the CBCL. High
heritability of the CBCL-JBD was demonstrated, with higher odds
ratios between monozygotic twins than between dizygotic twins
who fell into this latent class (Althoff et al 2006).
In this article, we explore the developmental stability of the
CBCL-JBD phenotype and determine the genetic and environ-
mental contributions to stability. One of the challenges to
understanding the prevalence and implication of child psycho-
pathology is the confound of development (Hudziak et al 2000).
Children, their brains and their behaviors, change over the
course of development. For instance, children with ADHD are
From the Department of Biological Psychology (DIB, IR, ED, TvB), Vrije Uni-
Harvard Medical School (RRA), Boston, Massachusetts; and the Depart-
ment of Psychiatry (DR, JJH), University of Vermont, Burlington, Ver-
Address reprint requests to Dr. Dorret Boomsma, Department of Biological
Psychology, Vrije Universiteit, Van der Boechorststraat 1, 1081 BT, Am-
sterdam, The Netherlands; E-mail: email@example.com.
Received July 21, 2005; revised February 10, 2006; accepted February 10,
BIOL PSYCHIATRY 2006;60:912–920
© 2006 Society of Biological Psychiatry
often less hyperactive as they grow up, and a certain percentage
of individuals no longer will suffer from ADHD as adults
(Mannuzza et al 2003). Similarly, aggression diminishes in boys
and girls over the course of development (Hudziak et al 2003;
Stanger et al 1997). In addition, genetic and environmental
influences on behavior may change with development. Changes
in genetic and environmental influences have been reported for
AP, AGG, and A/D behavior across development. Change across
development can be in the type of genetic and environmental
influences, as with the study of AP, in which the importance of
genetic dominance differs at different ages (Rietveld et al 2003a,
2003b, 2004). Changes can also occur in the magnitude of the
genetic and environmental influences, as our group reported
elsewhere for both AGG (Hudziak et al 2003) and A/D
(Boomsma et al 2005). By studying longitudinal twin data, it is
possible to determine the stability and change of behavioral
phenotypes across development and to assess the importance of
genetic influences to stability. The purpose of this article is to
assess the contribution of genetic and environmental factors to
the stability of the CBCL-JBD phenotype across ages 7, 10, and 12
years in longitudinal data from a large sample of more than 8000
twin pairs whose behavior was assessed by their mothers.
Methods and Materials
Subjects and Procedure
Data for this study come from an ongoing longitudinal study
that examines environmental and genetic influences on the
development of problem behavior in 3- to 12-year-old twins. The
families are volunteer members of the Netherlands Twin Register
(NTR), kept by the Department of Biological Psychology at the
Free University, Amsterdam (Boomsma et al 2002b). Starting in
1987, families with newborn twins were recruited. Currently,
40%–50% of all multiple births in The Netherlands are registered
by the NTR. For the present study, we included data of 7-, 10-,
and 12-year-old twin pairs (birth cohorts 1986 –1996). Parents of
twins were asked to fill in questionnaires about problem behav-
iors of the twins at ages 7, 10, and 12 years. After 2 months, a
reminder was sent to nonresponders. If finances permitted,
persistent nonresponders were contacted by phone. The total
sample consists of 8013 twin families. Appendix 1 shows the
distribution of the sample according to their participation across
time. There were 2866 families who participated three times, at
ages 7, 10, and 12 years; 2322 families participated twice, and
2825 families participated once. Most families who did not
participate at ages 10 and 12 years had not reached the proper
age for the twins (67% and 49%, respectively), and thus the
questionnaires have not been sent to them yet. Among those
who returned at least one survey at age 7, 10, or 12 years, 92%
returned the survey at age 7 years, 80%, at age 10 years, and 70%,
at age 12 years. If a family did not respond at a particular age,
they again were approached for the next mailing, so that
nonparticipants did not drop from data collection completely.
To examine the possible effect of sample attrition, data from
twins whose families participated three times were compared
with data from twins whose families participated at age 7 years
but did not return the questionnaires at ages 10 years, 12 years,
or both. The nonresponse group tended to show larger means in
JBD at age 7 (p ? .01). However, the Cohen’s effect size was less
than .20, which indicates a small effect. Furthermore, the possible
effects of sample attrition on the results of the present study are
minimized by inclusion of all available data in the analyses,
irrespective of the number of times that a family participated.
The sample includes 1331 MZM (monozygotic male twins),
1338 DZM (dizygotic male twins), 1537 MZF (monozygotic
female twins), 1250 DZF (dizygotic female twins), 1310 DOSMF
(dizygotic opposite-sex twins, male twin born first, female twin
born second), and 1240 DOSFM (dizygotic opposite-sex twins,
female twin born first and male twin born second). Zygosity
information was missing for seven pairs. For 1089 same-sex twin
pairs, zygosity was based on blood group (n ? 370) or deoxyri-
bonucleic acid (DNA; n ? 719) typing. For the remaining twins,
zygosity was determined by questionnaire items about physical
similarity and frequency of confusion of the twins by family and
strangers. Classification of zygosity was based on a discriminant
analysis of the questionnaire items and on zygosity based on
blood or DNA typing in same-sex twin pairs. The questionnaire-
based zygosity was correct for nearly 95% of the cases (Rietveld
et al 2000). A comparison of parental socioeconomic status (SES)
with the SES distribution for the general Dutch population
showed a slightly higher frequency of the middle and higher SES
groups (Rietveld et al 2003a). Representativeness of the sample
at each age is discussed by Van Beijsterveldt et al (2003). A
comparison of the twins’ emotional and behavioral problems at
age 3 year to that of singletons showed no differences between
twins and singletons (Van den Oord et al 1996).
At ages 7, 10, and 12 years, problem behavior was measured
with the CBCL/4-18 (Achenbach 1991). The CBCL consists of 118
items developed to assess behavioral and emotional problems.
Mothers were asked to rate the behavior of the child in the
preceding 6 months on a three-point scale. Children with more
than four missing items for the JBD phenotype were excluded
from the analyses. This occurred in fewer than 2.5% of the
CBCLs. The JBD phenotype was defined as the square-root
transformed sum of AP, AGG, and A/D.
Analyses were conducted by using structural equation mod-
eling, because it permits the simultaneous analysis of data from
multiple groups and allows imposition of parameter constraints
across groups. The statistical software packages Mx (Neale et al
2003) and Mplus (Muthen and Muthen 1998) were used. In
longitudinal studies such as the current one, not all subjects have
yet reached the oldest age, and not all subjects have taken part in
the study at all ages. To be able to use all data, full-information
maximum likelihood estimation with raw data was used. Twice
the negative log-likelihood (?2LL) of the data for each family is
calculated, and parameters are estimated so that the overall
likelihood of the raw data is maximized.
The fit of the genetic models is evaluated against the fit of a
saturated model, in which the covariance matrix and the mean
structures are computed without any restriction. Submodels were
compared with likelihood-ratio tests that are obtained by sub-
tracting ?2LL for a restricted nested model from that for a less
restricted model (?2? ??2LL0? ? ??2LL1?). The resulting test
statistic has a ?2distribution with degrees of freedom (df) equal
to the difference of the df between the two models. The ?2
statistic is sensitive to large sample sizes. Given large sample
sizes, small discrepancies between a model and the observed
data can lead to the rejection of the model (Loehlin 2004). The ?2
difference test applied to nested models has essentially the same
weaknesses as does the ?2test applied to any single model
(Schermelleh-Engle et al 2003). Thus, given the large sample size,
a confidence level of 99% (p ? .01) was chosen. In addition, we
D.I. Boomsma et al
BIOL PSYCHIATRY 2006;60:912–920 913
provide alternative goodness-of-fit measures such as the root
mean squared mean error of approximation (RMSEA) and
Akaike’s information criterion (AIC). The RMSEA is a measure of
closeness of fit and provides a measure of discrepancy per
degree of freedom. A value of .05 indicates a close fit, and values
up to .08 represent reasonable errors of approximation in the
population (Jöreskog 1993). The AIC compares models on the
basis of parsimony, taking jointly into account the ?2and the df
(Jöreskog 1993). The lower the AIC, the better the fit of the model
to the data, and the more parsimonious the model is.
The saturated model was used as a reference to test for the
homogeneity of means and variances. Homogeneity of means
and variances was tested, constraining them to be equal across
birth order, zygosity, sex, and time points (ages 7, 10, and 12 y).
For these tests, we also report the standardized root mean square
residual (SRMR) and the comparative fit index (CFI). The SRMR is
a badness-of-fit measure that is based on the fitted standardized
residuals; a value of zero indicates perfect fit, and values of less
than .10 may be interpreted as acceptable (Hu and Bentler 1995).
The saturated model has a SRMR of zero. Thus an increase in
SRMR is entirely a result of the specific homogeneity test and can
be seen as an indicator of the amount of variance that is
explained by heterogeneity. The CFI is a comparison index in
which the model of interest is compared with a baseline or
independence model. It is one of the fit indices that is less
affected by sample size and can take values from zero to one, for
which 97 or higher indicates a good fit, whereas values greater
than .95 may be interpreted as acceptable (Schermelleh-Engle
et al 2003).
The path diagram in Figure 1 represents the general genetic
model that was tested on the longitudinal JBD data. The diagram
represents the model for an opposite-sex twin pair. The first-born
twin is male, and the second-born twin is female. Different
parameters are estimated for male and female twins. The rectan-
gles represent the phenotypic measures at 7, 10, and 12 years for
both twins. A so-called ACE model was fitted in which the
variance of the JBD phenotype was explained by additive genetic
effects (A), environmental factors shared by the members of the
same family (C), and environmental factors specific to the
individual (E; the E component is omitted in Figure 1 for clarity).
The sources of variance are represented as latent, unmeasured
factors within circles. Genetic and environmental effects on
stability and change are investigated through a Cholesky or
triangular decomposition (Neale and Cardon 1992). Genetic (A)
and environmental (C and E) sources of variance-covariance
across time are represented by three latent factors, so that the first
factors are the stable sources of variance present at 7, 10, and 12
years of age; the second factors represent the sources of variance
common to 10 and 12 years of age that were not present at 7
years; and the third factors represent the sources of variance
specific to 12 years of age. That is, additive genetic effects are
represented by a triangular matrix of factor loadings, as follows:
a31 a32 a33?
(factors in columns and variables in rows); multiplying this
matrix by its transpose results in the genetic variance-covariance
matrix, as follows:
a31a11 a32a22?a31a21 a33
dividing this matrix by the implied phenotypic variance-covari-
ance matrix provides the proportion of variances and covari-
ances explained by additive genetic effects; and standardizing it
provides the genetic correlation matrix, in which the correlations
indicate the overlap of genetic effects across time.
Twins may resemble each other because they share their pre-
and postnatal rearing environment, often referred to as shared or
common environment (C). In addition, DZ twins may resemble
each other because they share 50% of their additive genetic
variance (A). MZ twins share all the additive genetic variance,
because they always, or nearly always, have identical genotypes.
Thus, A factors are correlated 1 across MZ twin pairs and .5
across DZ pairs. The correlation between genetic factors of OS
twins can be estimated (rgos), allowing for the possibility that
different genes influence the phenotype in male and female
twins. C factors are correlated 1 for MZ and DZ twins, and E
factors are uncorrelated between pairs by definition. Estimates of
the unique environmental effects (E) also include measurement
error (Boomsma et al 2002a). First, parameters in the full ACE
model were estimated. Next, equality constraints were imposed
across the sexes to test for sex differences in variance compo-
nents and were imposed across time to test for differences in
variance components across age.
Sample Characteristics and Descriptive Statistics
Table 1 shows the means and variances estimated in the
saturated model across zygosity and sex. Table 2 shows the results
of the tests for the homogeneity of means and variances. For all
tests, the SRMR and CFI indicate that only the mean and variance
differences across sexes can be considered relevant (SRMR ? .05,
CFI ? .97). There is a tendency for male twins to show larger means
and variability in CBCL-JBD than female twins.
The summary of twin correlations at each age and of the
cross-twin– cross-age correlations is shown in Table 3. The twin
correlations within age show that at each age, the DZ correlations
appear to be somewhat larger than half the MZ correlations. This
suggests that genes and shared family environment both explain
familial resemblances in CBCL-JBD. The cross-twin–cross-age
correlations represent JBD at one age (e.g., 7 y) in one twin, with
CBCL-JBD at another age (e.g., 10 y) in the other twin (correla-
tions constrained to be equal for first- with second-born twin and
for second-born with first-born twin). As can be seen, the past
behavior of the co-twin is more predictive for the current
behavior of his or her twin in MZ pairs than it is in DZ pairs. In
fact, for MZ twins, the cross-correlations are almost as high as
the within-person correlations across time. These within-person
correlations, or stability-coefficient correlations across time, were
.72 from 7 to 10 years, .66 from 7 to 12 years, and .77 from 10 to
12 years. On the basis of this pattern of cross-twin–cross-age
correlations for MZ and DZ twins, it may be expected that
longitudinal stability in JBD is explained by genetic factors and
by common environment.
Table 4 shows the standardized parameter estimates from the
full ACE model and from the reduced model without sex
differences. In these models, the estimate of the genetic correla-
914 BIOL PSYCHIATRY 2006;60:912–920
D.I. Boomsma et al
tion in opposite sex twins (rgos) was equal to .5, indicating that
the same genes are expressed in boys and girls. The fit of the
model was ?2LL ? 187,397.335, df ? 27849, ??2? 219.37,
?df ? 89, p ? .000, RMSEA ? .043 (.035–.049), AIC ? 41.37.
According to the RMSEA, the longitudinal model provides an
acceptable fit to the data. The estimates of the standardized variance
components (diagonals in Table 4) suggest that additive genetic
The off-diagonal estimates in Table 4 summarize the results
regarding the decomposition of the phenotypic stability across
time. The proportions above the diagonal give covariance com-
ponents, and the estimates below the diagonal give genetic and
environmental correlations across time. Genetic and environ-
mental covariance components sum to 100% and give the
BIP7_1 BIP10_1 BIP12_1
BIP7_2 BIP10_2 BIP12_2
Figure 1. Analysis of longitudinal data on juvenile bipolar disorder (JBD) at ages 7, 10, and 12 years (nonshared environment is omitted in the figure for
simplicity but is modeled in a similar way). The figure shows data from a pair of dizygotic opposite-sex twins; the rectangles represent the phenotypic
family (C). These are represented as latent, unmeasured factors within circles. Genes (A) and environment (C) across time are represented by three latent
0.5 in same-sex dizygotic twins and may be estimated in opposite-sex dizygotic twins (rgos).
D.I. Boomsma et al
BIOL PSYCHIATRY 2006;60:912–920 915
proportion of the total covariance across time that is explained
by genetic and environmental stable influences. For CBCL-JBD,
results suggest that roughly 80% of the stability in childhood is a
result of additive genetic effects, and about 10% of stability from
10 to 12 years of age is explained by shared environmental
influences. Genetic and environmental correlations can be inter-
preted as indicators of the extent to which the same genes and
environmental factors influence the trait at different ages. As may
be seen, genetic correlations are high (.7 or above), whereas
environmental correlations are lower.
Table 5 summarizes the model-fitting tests. Model 1 is the ACE
model, with sex differences in parameter estimates. In model 2,
the factor loadings of the A, C, and E latent factors are con-
strained to be equal for male and female twins. Model two fits the
data significantly worse than does model 1. When the absolute
amount of variance explained by each component was con-
strained independently, only the C component could be con-
strained to be equal for male and female twins (??2? 11.239,
?df ? 6, p ? .081), whereas the amount of variance explained by
A and E differed significantly across sexes (p ? .01).
The results concerning sex differences may appear surprising
given that the estimates of the proportion of variance explained
by A, C, and E in the full ACE model (Table 4) do not look so
different. The explanation might rest in the constraint that the
absolute factor loadings are equal across sexes, for example, as
in A: am11? af11, am21? af21, am31? af31, am22? af22, am32?
af32, and am33? af33.However, the proportion of variance that is
explained by each component also depends on the total vari-
ance. Thus, although the proportion of variance explained by A,
C, and E was equal for male and female twins, the absolute
amounts of variance explained was larger for male twins who
have larger total variances. To allow for this possibility, nonlinear
constraints were used to test whether the relative proportion of
variance accounted for by A, C, and E was equal across sexes, for
example, as in A:
In model 3, only the cross-sectional variance components are
constrained across sexes as shown in the previous equations.
Then, in model 4, the same constraint is extended to the
cross-time A, C, and E covariance components. Model 4, in which
both relative variance and covariance components are con-
strained to be equal across sexes, fits the data as well as the full
ACE model, and when compared with the saturated model, it
presents an RMSEA of .013 (.011–.015), indicative of an excellent
fit. Thus, it can be concluded that the same relative amount of
Table 1. Means and Variances of CBCL-JBD Across Zygosity by Sex, at Ages 7, 10, and 12 Years
MeanVariancen (Twin Pairs)
7 y10 y12 y7 y10 y12 y7 y10 y
MZM, monozygotic males; DZM, dizygotic males; MZF, monozygotic females; DZF, dizygotic females; DOS, dizygotic opposite-sex pairs.
Table 2. Tests for Homogeneity of Means and Variances
Homogeneity of means
Across birth order
Age 7 ? age 10
Age 10 ? age 12
Age 7 ? age 10 ? age 12
Homogeneity of variances
Across birth order
Across age 7 ? 10 ? 12
??2, Change in chi-squared statistic and degrees of freedom (df) com-
pared with to a fully saturated model; SRMR, standardized root mean
squared residual index; CFI, comparative fit index.
aP ? .01.
Table 3. Twin Correlations at Ages 7, 10, and 12 Years and
Cross-Twin-Cross-Time Correlations for JBD
MZM, monozygotic males; DZM, dizygotic males; MZF, monozygotic
females; DZF, dizygotic females; DOS, dizygotic opposite sex pairs.
Confidence intervals for correlations are given in Appendix 2.
916 BIOL PSYCHIATRY 2006;60:912–920
D.I. Boomsma et al
variance within time and covariance across time is explained by
A, C, and E for male and female twins.
Table 4 shows the parameter estimates of model 4, in which
standardized estimates are the same for girls and boys.
Between ages 7 and 12 years, the heritability of JBD increases
from 63% to 75%, and the contribution of shared environment
decreases from 20% to 8%. The remaining 18% of the variance
is explained by unique environment. Covariance components
show that the largest part of the stability (between 75% and
84%) between 7 and 12 years is a result of additive genetic
effects. Only 4% and 15% is a result of shared environment,
and around 10% is a result of stable unique environmental
Finally, models 5, 6, and 7 tested differences in variance
explained by A, C, and E across time. Nonlinear constraints were
used to test whether the variance explained by A, C, and E was
proportional at ages 7, 10, and 12 years; that is, for A,
Models 5 and 6 fit significantly worse than model 4, whereas
model 7 fit as well as model 4. According to these results,
additive genetic effects increase significantly with age, whereas
the effects of the shared environment decrease. The proportion
of variance explained by the unique environment remains the
Finally, given that the amount of variance explained by C
decreases to values close to zero at age 12 years, three additional
models were fitted in which the C component was constrained to
be zero at ages 12 years (c33? c32? c31? 0), 10 years (c22?
Table 4. Standardized Parameter Estimates from the ACE Model with/without Sex Differences
Parameter Estimates from Full ACE Model
Parameter Estimates from
Reduced Model Without
Sex Differences MalesFemales
7 10127 10 12710 12
Additive genetic architecture (heritability on diagonal, genetic covariance components above and genetic
correlations below diagonal)
Shared environment architecture (% of variance explained by shared environment on diagonal, shared
environmental covariance components above and correlations below diagonal)
.47 .29 .08.65
Unique environment architecture (% of variance explained by unique environment on diagonal, unique
environmental covariance components above and correlations below diagonal)
.37.15 .12 .46
.35 .57 .17.39
Confidence intervals are shown in Appendix 3.
Table 5. Model Fitting Results: Tests for Sex Differences and Longitudinal Trends Based on the ACE Model (Additive Genetic, Common Environmental and
Unique Environmental Influences)
Tests for sex differences in absolute estimates of variance components
1 Full ACE with rg OS ? .5
2 No sex differences in variance components
Tests for sex differences in standardized variance components and in longitudinal covariance components
3 Proportion of Variance explained by ACE equal for males and females
4 Proportion covariance explained by ACE equal for males and females
Tests of longitudinal changes in the proportion of variance explained by A, C and E
5 Proportion of variance explained by A equal at 7, 10 and 12 years
6 Proportion of variance explained by C equal at 7, 10 and 12 years
7 Proportion of variance explained by E equal at 7, 10 and 12 years
aC.T. Compared to model number #.
D.I. Boomsma et al
BIOL PSYCHIATRY 2006;60:912–920 917
c21? 0), and 7 (c11? 0) years. The results showed that the effects
of C are significant at ages 7 and 10 years (p ? .01) but that they
are negligible at age 12 years (p ? .119).
This study examined the stability and genetic architecture
across time of the CBCL-JBD phenotype, which has been shown
to be consistent with DSM conceptualizations of JBD. CBCL-JBD,
defined as the sum of the AP, AGG, and A/D subscales, has been
demonstrated to be associated with the DSM JBD phenotype
across studies. The use of CBCL-JBD as a measure of DSM JBD
has been recommended as one possible method to circumvent
the diagnostic confounds that continue to be debated (National
Institute of Mental Health 2001). The CBCL-JBD construct has the
advantage that it is based on empirically derived dimensions of
childhood psychopathology, whose summation leads to a con-
tinuous scale. The use of continuous scales, as compared with
categorical or dichotomous data, leads to an increase in statistical
power in genetic studies (e.g., Derks et al 2004; Neale et al 1994).
By using the summed score, children with DSM-III-R BD were
identified accurately in a large family study population (Faraone
et al, in press).
We have found that the CBCL-JBD measure is stable across
ages, and we have quantified the genetic and environmental
contributions to the variation of CBCL-JBD and to its stability
from ages 7 to 12 years. The influence of additive genetic effects
on variation in JBD was found to be relatively high at each age,
increasing from 63% at age 7 years to 75% at age 12 years.
Simultaneously, the effects of the shared environment tend to
decrease. At age 7 years, 20% of the variation in CBCL-JBD is
explained by the influence of the common family environment,
and this percentage decreases to 8% at age 12 years. The small
remaining part of the variance at each age was explained by
unique or individual-specific environmental influences. The esti-
mates of common and unique environmental variances may be
somewhat biased. The CBCL-JBD scale shows a skewed distribu-
tion. When an ACE model is fitted to such data, an unbiased
estimate of the additive genetic effect is obtained (Derks et al 2004).
However, the common environmental effect may be underesti-
mated at the cost of the unique environmental effect.
The standardized estimates of genetic and environmental
parameters were found to be the same for boys and girls, but not
across time. The heritability estimates were 63%, 71%, and 75% at
ages 7, 10, and 12 years, respectively. Our analyses also suggest
that similar genes may underlie CBCL-JBD for both girls and
boys. The estimates for the percentage of variance explained by
common family environment were 20%, 11%, and 8%. Overall,
the high heritability estimates obtained for this sample are in line
with those obtained in adults with BD (see Smoller and Finn 2003
for a review).
Another major finding from this study concerned the stability
of the CBCL-JBD phenotype across development. Correlations
across age groups were .72 from 7 to 10 years, .66 from 7 to 12
years, and .77 from 10 to 12 years. Genetic covariance analysis
suggests that roughly 80% of the stability on JBD in childhood is
a result of additive genetic effects and that about 10% of stability
is a result of shared environmental effects. Should this finding
hold into adulthood (and, as important, should the phenotypic
association between adult-onset BD and JBD be delineated
further), it would suggest that many of the candidate chromo-
somal locations for adult BD genes, including 6q16–22 and
12q23–24 among several others (Boomsma et al 2006; Craddock
et al 2005; Dick et al 2003), also may be important in CBCL-JBD.
However, regardless of the association between JBD and adult-
onset BD, research into the genetic influences on differences
among children in JBD has merit in its own right. The
association between JBD and adult-onset BD has yet to be
clearly established and is considered an important research
topic by the leaders in this field (e.g., National Institute of
Mental Health 2001). It is possible that early-onset forms of BD
have fundamentally different and developmentally important
genetic and gene by environment effects than the adult-onset
form. These effects cannot be studied without large, longitu-
dinal samples. This is an important topic for further research
and is a future aim of our work.
Interestingly, the influences of common environment appeared
to decrease over time, particularly between ages 7 and 10 years.
Putting together this result with the overall decrease in influence of
shared environment from 7 to 12 years suggests the possibility of an
important environmental factor during the formative years of CBCL-
JBD onset that may not be present later on. Alternatively, this effect
could include rater bias (e.g., stereotyping or having certain re-
sponse styles). Rater bias in this sense will be a continuous process
influencing the ratings at all ages and could mimic stability in the
trait. Maternal psychopathology is one example that could affect
ratings of problem behavior in their children. Because rater bias
affects MZ and DZ twin correlations in the same way, it will appear
as shared environmental effects. Also, assortative mating in parents
could appear as a shared environmental effect. However, for both
phenomena, we probably would not expect that their effects
diminish between 7 and 12 years.
These findings for CBCL-JBD are in contrast with those for the
separate subscales. Modeling of the AP phenotype across ages
3–12 showed additive and dominance genetic effects, along with
unique environmental effects and no common environmental
effects (Rietveld et al 2003a, 2004). This was replicated in a study
comparing the CBCL-JBD profile with the CBCL-AP profile
(Hudziak et al, in press). It appears therefore, that the CBCL-JBD
is different in terms of its heritability with AP and is unlikely to be
an extreme form of that phenotype.
Geller and colleagues (2001) have shown that there is a high
degree of overlap between childhood-onset major depression
and adult diagnosis of BD. Could CBCL-JBD be an expression of
A/D? When we look at the previous modeling of A/D, this
explanation appears unlikely. In our work on A/D, we found that
although additive genetic, common, and unique environmental
factors are important (similar to CBCL-JBD), the heritability of
A/D decreases with increasing age (from ages 3 to 12 y), with the
common environmental component increasing—exactly the op-
posite pattern that is seen with CBCL-JBD (Boomsma et al, 2006).
Thus, although the shared family environment becomes more
important to the expression of A/D as the child ages, it becomes
less important to the expression of CBCL-JBD. We have demon-
strated similar increases in the contribution of the shared envi-
ronmental factor in AGG (Van Beijsterveldt et al 2003) but only in
female twins. Male twins showed a relatively consistent contri-
bution of shared environmental contribution for AGG across
childhood. Overall, these findings suggest that the CBCL-JBD
construct is something different than its component parts.
In summary, this study provides evidence from a large sample
that many of the symptoms comprising JBD are stable across time
and are strongly influenced by additive genetic factors that tend
to increase with time in contrast to shared environmental factors
which tend to decrease. Moreover, this stability of the CBCL-JBD
phenotype also is due in large part to additive genetic influences.
918 BIOL PSYCHIATRY 2006;60:912–920
D.I. Boomsma et al
It is important to note that we observed no sex differences in
genetic architecture or in the stability of the CBCL-JBD pheno-
type, indicating that for gene-finding studies, data may be pooled
across boys and girls.
This work was supported by National Institute of Mental
Health Grant No. MH58799 and Nederlandse Organisatie voor
Wetenschappelijk Onderzoek Grant Nos. 575-25-006, 575-25-
012, and 904-57-94.
Profile. Burlington, VT: University of Vermont Department of Psychiatry.
Althoff RR, Faraone SV, Rettew DC, Morley CP, Hudziak JJ (2005): Family,
twin, adoption and molecular genetic studies of juvenile bipolar disor-
der. Bipolar Disord 7:598–609.
Althoff RR, Rettew DC, Faraone SV, Boomsma DI, Hudziak JJ (2006): Latent
notype. Biol Psychiatry 60:903–911.
for ADHD in prepubertal children. J Am Acad Child Adolesc Psychiatry
clinical scales discriminate prepubertal children with structured inter-
mental influences on anxious/depressed behavior during childhood: A
Boomsma D, Busjahn A, Peltonen L (2002a): Classical twin studies and be-
Boomsma DI, Cacioppo JT, Slagboom PE, Posthuma D (2006): Genetic link-
age and association analysis for loneliness in Dutch twin and sibling
Carlson GA, Bromet EJ, Sievers S (2000): Phenomenology and outcome of
subjects with early- and adult-onset psychotic mania. Am J Psychiatry
Carlson GA, Kelly KL (1998): Manic symptoms in psychiatrically hospitalized
children—What do they mean? J Affect Disord 51:123–135.
and bipolar disorder: Dissecting psychosis. J Med Genet 42:193–204.
estimates and power in genetic modeling. Twin Res 7:659–669.
Dick DM, Foroud T, Flury L, Bowman ES, Miller MJ, Rau NL, et al (2003):
Genomewide linkage analyses of bipolar disorder: A new sample of 250
pedigrees from the National Institute of Mental Health Genetics Initia-
Dienes KA, Chang KD, Blasey CM, Adleman NE, Steiner H (2002): Character-
ization of children of bipolar parents by parent report CBCL. J Psychiatr
Faedda GL, Baldessarini RJ, Glovinsky IP, Austin NB (2004): Pediatric bipolar
disorder: Phenomenology and course of illness. Bipolar Disord 6:305–
Faedda GL, Baldessarini RJ, Suppes T, Tondo L, Becker I, Lipschitz DS (1995):
Pediatric-onset bipolar disorder: A neglected clinical and public health
Faraone SV, Althoff RR, Hudziak JJ, Monuteaux M, Biederman J (2005): The
CBCL predicts DSM bipolar disorder in children: A receiver operating
characteristic curve analysis. Bipolar Disord 7:518–524.
Geller B, Warner K, Williams M, Zimerman B (1998): Prepubertal and young
adolescent bipolarity versus ADHD assessment and validity using the
WASH-U-KSADS, CBCL and TRF. J Affect Disord 51:93–100.
Geller B, Zimerman B, Williams M, Bolhofner K, Craney JL (2001): Bipolar
disorder at prospective follow-up of adults who had prepubertal major
Hazell PL, Lewin TJ, Carr VJ (1999): Confirmation that Child Behavior Check-
list clinical scales discriminate juvenile mania from attention deficit
Hu L, Bentler P (1995): Evaluating model fit. In: Hoyle RH, editor. Structural
Hudziak JJ, Althoff RR, Derks EM, Faraone SV, Boomsma DI (2005): Preva-
lar disorder. Biol Psychiatry 58:562–568.
inattentive, aggressive, and anxious/depressed behaviors. J Am Acad
editors. Testing Structural Equation Models. Newbury Park, CA: Sage,
Kahana SY, Youngstrom EA, Findling RL, Calabrese JR (2003): Employing
parent, teacher, and youth self-report checklists in identifying pediatric
bipolar spectrum disorders: an examination of diagnostic accuracy and
Leibenluft E, Charney DS, Towbin KE, Bhangoo RK, Pine DS (2003): Defining
clinical phenotypes of juvenile mania. Am J Psychiatry 160:430–437.
Loehlin J (2004): Latent Variable Models, 4th ed. Mahwah, NJ: Erlbaum. Man-
hyperactivity disorder into adulthood: What have we learned from the
prospective follow-up studies? J Atten Disord 7:93–100.
National Institute of Mental Health (2001): National Institute of Mental
Health research roundtable on prepubertal bipolar disorder. J Am Acad
Neale M, Boker SM, Xie G, Maes H (2003): Mx: Statistical Modeling, 6th ed.
Virginia Institute for Psychiatric and Behavioral Genetics, VCU Box
900126, Richmond, VA 23298: Department of Psychiatry.
Neale M, Cardon L (1992): Methodology for Genetic Studies of Twins and
Families North Atlantic Treaty Organization, Scientific Affairs Division.
Dordrecht, The Netherlands: Kluwer Academic.
to resolve variation in threshold traits. Behav Genet 24:239–258.
Heritability of attention problems in children: I. Cross-sectional results
Rietveld MJH, Hudziak JJ, Bartels M, van Beijsterveldt CEM, Boomsma DI
(2004): Heritability of attention problems in children: Longitudinal
results from a study of twins, age 3 to 12. J Child Psychol Psychiatry
Rietveld MJ, Posthuma D, Dolan CV, Boomsma DL (2003b): ADHD: Sibling
interaction or dominance: An evaluation of statistical power. Behav
(2000): Zygosity diagnosis in young twins by parental report. Twin Res
Schermelleh-Engle K, Moosbrugger H, Muller H (2003): Evaluating the fit of
structural equation models: Tests of significance and descriptive good-
Smoller JW, Finn CT (2003): Family, twin, and adoption studies of bipolar
Stanger C, Achenbach TM, Verhulst FC (1997): Accelerated longitudinal
comparisons of aggressive versus delinquent syndromes. Dev Psycho-
Van Beijsterveldt CEM, Bartels M, Hudziak JJ, Boomsma DI (2003): Causes of
nal genetic analysis in Dutch twins. Behav Genet 33:591–605.
Van den Oord EJ, Verhulst FC, Boomsma DI (1996): A genetic study of
Wals M, Hillegers MH, Reichart CG, Ormel J, Nolen WA, Verhulst FC (2001):
D.I. Boomsma et al
BIOL PSYCHIATRY 2006;60:912–920 919
Weller EB, Weller RA, Fristad MA (1995): Bipolar disorder in children: Misdi-
agnosis, underdiagnosis, and future directions. J Am Acad Child Adolesc
Appendix 1. Participation Rates by Age
7 y10 y No ResponseResponseNo Questionnaire SentTotal
No responseNo response
No questionnaire sent
No questionnaire sent
No questionnaire sent
No questionnaire sent1
Appendix 2. 99% Confidence Intervals for Twin Correlations
Cross Twin–Within Time (by Ages in y)Cross Twin–Cross Time (by Ages in y)
7 10 127–10 7–1210–12
Appendix 3. 99% Confidence Intervals for Model 4 (Reduced Model)
Parameter Estimates from Model without
Age in y7 y 10 y12 y
Additive genetic architecture (heritability on diagonal, with genetic
covariance components above and genetic correlations
Shared environment architecture (% of variance explained by shared
environment on diagonal, with shared environmental covariance
components above and correlations below diagonal)
Unique environment architecture (% of variance explained by unique
environment on diagonal, with unique environmental covariance
components above and correlations below diagonal)
920 BIOL PSYCHIATRY 2006;60:912–920
D.I. Boomsma et al