Arch Womens Ment Health (2006) 9: 181–186
Identifying the behavioural phenotype in fetal alcohol spectrum
disorder: sensitivity, specificity and screening potential
K. Nash1, J. Rovet1;3, R. Greenbaum4, E. Fantus1, I. Nulman2;3, and G. Koren2;3
1Psychology Department, The Hospital for Sick Children, Toronto, Cananda
2Motherisk Program, The Hospital for Sick Children, Toronto, Cananda
3Department of Pediatrics, University of Toronto, Toronto, Cananda
4The Hincks-Dellcrest Centre, Toronto, Cananda
Received July 21, 2005; accepted March 11, 2006
Published online May 3, 2006 # Springer-Verlag 2006
Background: In most cases of Fetal Alcohol Spectrum Disorder
(FASD), the pathognomonic facial features are absent making diagno-
sis challenging, if not impossible, particularly when no history of
maternal drinking is available. Also because FASD is often comorbid
with Attention Deficit Hyperactivity Disorder (ADHD), children with
FASD are frequently improperly diagnosed and receive the wrong
treatment. Since access to psychological testing is typically limited
or non-existent in remote areas, other diagnostic methods are needed to
provide necessary interventions.
Objectives: To determine if a characteristic behavioural phenotype
distinguishes children with FASD from typically developing children
and children with ADHD and use this information to create a screening
tool for FASD diagnosis.
Methods: Parents and caregivers completed the Child Behavior
Checklist (CBCL), a well-established standardized tool for evaluating
children’s behavioural problems. Results from 30 children with Fetal
Alcohol Syndrome or Alcohol-Related Neurodevelopmental Disability,
30 children with ADHD, and 30 typically developing healthy children
matched for age and socioeconomic status with FASD were analyzed.
Based on our previous work, 12 CBCL items that significantly differ-
entiated FASD and control groups were selected for further analyses.
Stepwise discriminant function analysis identified behavioural charac-
teristics most strongly differentiating groups and Receiver Operating
Characteristics (ROC) curve analyses determined sensitivity and
specificity of different item combinations.
Results: Seven items reflecting hyperactivity, inattention, lying and
cheating, lack of guilt, and disobedience significantly differentiated
children with FASD from controls. ROC analyses showed scores of
6 or higher on these items differentiated groups with a sensitivity of
86%, specificity of 82%. For FASD and ADHD, two combinations
of items significantly differentiated groups with high sensitivity and
specificity (i) no guilt, cruelty, and acts young (sensitivity¼70%;
specificity¼80% (ii) acts young, cruelty, no guilt, lying or
cheating, steals from home, and steals outside (sensitivity¼81%;
specificity¼72%). These items were used to construct a potential
FASD screening tool.
Conclusions: Our findings identifying the behavioural characteristics
differentiating children with FASD from typically developing children
or children with ADHD have the potential for development of an
empirically derived tool for FASD tool to be used in remote areas
where psychological services are not readily available. This technique
may speed up diagnosis and intervention for children without ready
access to formal assessments.
Keywords: FASD; ADHD; screening; child behavior checklist;
alcohal; ethanol; pregnancy.
The damaging effects of prenatal alcohol exposure for
human development are well recognized and lead to a
condition known as Fetal Alcohol Spectrum Disorder
(FASD). FASD involves several conditions resulting
from gestational exposure to alcohol with the two
most common being Fetal Alcohol Syndrome (FAS)
and Alcohol-Related Neurodevelopmental Disability
(ARND; Jones & Smith, 1973; Stratton et al, 1996).
These conditions give rise to a number of challenges
including mild to severe behavioural disturbance and
significant developmental delay making FASD the single
most prevalent preventable cause of congenital neurobe-
havioural dysfunction in the Western world (Carmichael
et al, 1998; Mattson & Riley, 1998; Sampson et al,
1997). Because a large proportion of individuals with
FASD require extensive mental health services through-
out their lifetime, the costs associated with FASD are
staggering. Indeed, it is estimated that in Canada $344
million are spent annually on affected youth while in the
United States, researchers estimate that a single child
with FAS=ARND requires 1.4 million dollars of inter-
vention in his or her lifetime (Stade et al, 2003). Since
incarceration and difficult-to-measure costs such as lost
productivity and poor quality of life are excluded from
these estimates, the actual price of FASD is likely much
higher. It is recognized that early intervention can help
alleviate some of the debilitating consequences of FASD,
therefore proper diagnosis must be made (Streissguth,
For FAS, the diagnostic criteria include a character-
istic facial dysmorphology, pre- and postnatal growth
retardation, a complex and pervasive pattern of neuro-
behavioural anomalies, and a history of maternal drink-
ing during pregnancy (Chudley et al, 2005; Astely,
2004). In ARND, by contrast, which affects as many
as 90% of cases, the distinctive facial features and
growth abnormalities are typically absent necessitating
ic neurobehavioural features (Greenbaum et al, 2002;
Greenbaum, 2004). However, because many children
with ARND are adopted or fostered, information about
their biological mothers’ drinking patterns during preg-
nancy is often not available. Furthermore, since proper
neurobehavioural diagnosis requires a specialized team
of professionals including psychometrists, psycholo-
gists, and pediatricians, diagnosis is costly, involves long
wait lists, and does not adequately serve children living
in remote areas where access to full diagnostic services
and mental health professionals is limited. Because
FAS=ARND is also comorbid with other psychiatric dis-
orders, such as Attention Deficit Hyperactivity Disorder
(ADHD), occurring in as many as 70% of FASD chil-
dren, children with ARND are typically improperly
may not be sufficient to ameliorate all aspects of FASD
(Nanson &Hiscock, 1990; Coles etal,1997; Greenbaum,
2004). Thus, there is an urgent need to develop better
methods for identifying children with FASD, one that is
both effective in discriminating children with FASD
from ADHD and that can be used in remote areas.
Since its inception in 1996, the Motherisk FAS clinic
has conducted comprehensive evaluations on over 200
children in order to diagnose FAS or ARND according to
both Institute of Medicine criteria and our own empiri-
cally derived standards (Greenbaum et al, 2001; Astley,
2004) for ruling out non-cases from cases. During this
evaluation, parents or caregivers completed the Child
Behaviour Checklist (CBCL), a well-established stan-
dardized questionnaire for assessing behaviour problems
in children (Achenbach & Rescorla, 2001). This instru-
ment serves to identify problematic areas of behaviour
(e.g., attention problems) as well as resemblance to clin-
ical psychiatric syndromes (e.g., ADHD). Recent studies
have applied advanced statistical techniques to CBCL
results in order to identify the characteristic behavioural
phenotypes in selective pathological conditions. How-
ever, this research was based on the measure’s global
scales and not the specific behaviours (Hudziak et al,
2004; Biederman et al, 2005). We have previously shown
that individual CBCL items can be effectively used to
distinguish children with FASD from normal controls
and significant characteristics unique to FASD, thereby
constituting a distinct phenotype (Greenbaum, 2000).
Also, because the particular items can be studied apart
from the test as a whole, they have potential to be used as
a screening tool in diagnosing FAS or ARND, especially
in remote areas where access to testing is limited.
Thus the purpose of the present study was to replicate
our earlier findings using a newer version of the CBCL
and determine the particular item combinations best dis-
criminating FASD and ADHD groups. A supplementary
goal was to use this information to develop an empiri-
cally derived FASD screening tool.
The FASD group consisted of 54 children with FAS (n¼11) or
ARND (n¼43) diagnosed in the Motherisk FAS clinic. They
were drawn from a sample of 75 cases aged 6 to 16 years who
had received a diagnosis of FAS or ARND and whose parent or
caregivers had completed the CBCL. All children were brought
to the clinic by (a) foster or adoptive parent regarding concerns
about suspected alcohol exposure or (b) biological relative
concerned whether child’s current problems reflected his or
her prenatal alcohol exposure (Greenbaum et al, 2002). In all
cases, exposure history was confirmed by one of three criteria:
(i) verbal report of the biological parent or relative, (ii) the child
suffered alcohol withdrawal at birth, or (iii) the child was placed
in care because of maternal alcohol abuse. All children were
evaluated by both a physician who measured their height and
used a validated system for evaluating facial dysmorphology and
by a team of psychologists and psychometrists who adminis-
tered a battery of neuropsychological tests to each child. All
parents or guardians provided signed informed consent allowing
us to use clinical information for research purposes.
Controls were typically developing 6 to16 year old children
recruited from local schools in the Toronto area, postings within
the hospital and who had participated in previous studies
K. Nash et al
(Greenbaum, 2004; Hepworth, 2004). Controls were screened
for learning disabilities, for ADHD and maternal history of
alcohol consumption in pregnancy using parent-completed
Children with ADHD were recruited from the practices of
local behavioural pediatricians or psychiatrists and by advertis-
ing at a local ADHD parent support group. To be included, all
children had to have received a DSM-IV diagnosis of ADHD.
Any child with a report of maternal drinking during pregnancy
From the 54 potential cases, 30 (5 FAS & 25 FAE) were
suitable for age matching with children with ADHD. This
research was approved by the Research Ethics Board at The
Hospital for Sick Children.
Test and analytic methods
The Child Behavior Checklist (CBCL; Achenbach & Rescorla,
2001) is a parent=caregiver questionnaire, which assesses social
competencies and behaviour problems in children aged 6 to 18
years. The CBCL is comprised of both a series of open-ended
questions and a rating scale of 113 behavioural descriptors
scored on a 3-point scale from 0¼not true, 1¼sometimes true,
and 2¼often true. Computer scoring of the CBCLyields a Total
Behaviour Problems score, two broad-band scores assessing
Internalizing and Externalizing behaviour problems, and eight
narrow-band scales assessing Withdrawn=Depressed, Somatic
Complaints, Anxious=Depressed, Social, Thought, Attention,
Rule-Breaking Behaviour, and Aggressive problems.
Our previous work comparing 35 children with ARND and
35 controls (Greenbaum, 2000) matched for age, gender and
SES on all 113 CBCL items showed significant differences on
62 items with 12 differing beyond the p<0.001 level. These
12 items were: ‘acts too young for age’, ‘argues’, ‘can’t con-
centrate=poor attention’, ‘can’t sit still=restless=hyperactive’,
‘cruelty, bullying or meanness to others’, ‘disobedient at home’,
‘no guilt after misbehaving’, ‘impulsive=acts without thinking’,
‘lying or cheating’, ‘showing off=clowning’, ‘steals from home’,
and ‘steals outside’. In the present study, we selected these 12
items for comparing the three groups of children. Items were
currently scored 1 or 0 depending on whether they were
endorsed (with a 1- or 2-point score) or not respectively.
Statistics and data analyses
Item frequencies were compared among groups using chi-square
analyses. Discriminant function analyses (DFA) conducted sepa-
rately between the FAS=ARND group and controls and between
FAS=ARND and ADHD groups served to identify items most
strongly differentiating the groups. Finally, Receiver Operating
Characteristic (ROC) curve analyses were performed on differ-
ent combinations of frequently endorsed items to determine
sensitivity, specificity, and cut-off scores. Area-under-the-curve
(AUC) values were used to classify cases as being FAS=ARND
or not based on the number of endorsed items and critical cutoff
Table 1 shows the endorsement rates for each group on
all 12 items. Children with FAS=ARND had significantly
higher endorsement rates than controls on every item with
all but two differing at the p<0.001 level. Children with
FAS=ARND had higher endorsement rates than ADHD
on6 ofthe 12 items, particularly‘‘acts young’’,‘‘cruelty’’,
‘‘no guilt’’, ‘‘lying or cheating’’, ‘‘stealing from home’’,
and ‘‘stealing from outside’’. Children with ADHD had
higher endorsement rates than controls on all items except
‘‘cruelty’’. Although not significant, children with ADHD
were slightly more restless than FASD.
Discriminant function analysis was used to identify the
items and their order of predicting differences between
groups. A comparison between children with FAS=ARND
and controls revealed the following set of 7 items
most strongly differentiated these groups (?2(7)¼72.2,
p<0.001): ‘‘no guilt’’, ‘‘lying or cheating’’, ‘‘can’t
Table 1. Endorsement rates for FASD, ADHD, and control groups on the 12 discriminating CBCL items
Item descriptionPercent endorsements p-value
FASDADHD ControlFASD vs
Acts young for age
Argues a lot
Can’t concentrate or poor attention
Can’t sit still, restless, hyperactive
Cruelty, bullying or meanness to others
Disobedient at home
Doesn’t show guilt after misbehaving
Impulsive or acts without thinking
Lying or cheating
Showing off or clowning
Steals from home
Behavioural phenotype in FASD
concentrate’’, ‘‘restless’’, ‘‘impulsive’’, ‘‘disobedient’’,
‘‘acts young’’. Table 2 shows the canonical correlations
for these 7 items. A similar comparison between FAS=
ARND and ADHD groups revealed a different combina-
tion of discriminating items (?2(6)¼38.4, p<0.001):
‘‘steals outside’’, and ‘‘lying or cheating’’.
Next, we submitted different combinations of the
most discriminating items to ROC curve analyses. Com-
parison between FAE=ARND children and controls
revealed the largest AUC was achieved with the full
set of 7 items as shown in Table 2 (AUC¼0.965,
p<0.001). With a cutoff of 6 (i.e., selecting 6 of the 7
items for a child), a sensitivity of 86% and specificity of
82% could be achieved. The next strongest combination
of items were the following five, which produced an
AUC of 0.901 (p<0.001) with a cutoff of 3 yielding
maximum sensitivity and specificity of 70% and 80%
respectively: ‘‘no guilt’’, ‘‘lying=cheating’’, ‘‘disobedient
at home’’, and ‘‘acts young’’.
ROC curve analyses comparing FAS=ARND and
ADHD groups yielded an AUC of 0.863 (p<0.001)
with the full set of 6 items differentiating these groups
(see Table 3). A cutoff of 3 items yielded sensitivity of
81% and specificity of 72% (see Fig. 2). Surprisingly,
the next largest combination of items consisting of ‘‘no
guilt’’, ‘‘cruelty’’, and ‘‘acts young’’ yielded almost as
large an AUC as with 6 items (0.849, p<0.001) with
endorsements of 2 items yielding a sensitivity of 70%
and specificity of 80%.
Present findings based on analysis of the 12 CBCL items
we previously found to most highly discriminate chil-
dren with ARND from controls showed that these behav-
iours continue to differentiate strongly children with
FASD and typically developing children as well as chil-
dren with ADHD. These findings are of paramount
importance because these particular items can lay the
groundwork for developing a screening tool, which can
accurately and consistently distinguish FAS=ARND
Table 2. Canonical correlation coefficients for items shown to discri-
minate FASD and control groups on the discriminant function analysis
Order Item Canonical correlation
lying or cheating
Table 3. Canonical correlation coefficients for items shown to discri-
minate FASD and ADHD groups on the discriminant function analysis
Order ItemCanonical correlation
steals from home
lying or cheating
Fig. 2. ROC curves showing items most significantly discriminating
children with FASD from ADHD.
All 6.No guilt, cruelty,
Fig. 1. ROC curves showing items most significantly discriminating
children with FASD from Controls.
cheating, disobedient at home, acts young
All 7.No guilt, lying,
K. Nash et al
from ADHD and lead to earlier interventions for chil-
dren with FASD, particularly those who live in remote
areas where access to diagnostic services are not often
available. We found that while children with FASD
exhibited attention deficits and hyperactivity, as do chil-
dren with ADHD; the FASD group unlike the ADHD
group additionally displayed a lack of guilt after mis-
behaving, cruelty, and a tendency to act young for their
age. Furthermore, children with FASD were also more
likely to lie and steal than children with ADHD.
The prevalence of FASD estimated to be about 0.91
percent in the general population is as high as 10 to 20
1997; Sokol et al, 2003). Since many of these com-
munities are isolated, these children with FASD may not
have access to a psychologist. This is especially proble-
matic because the majority of children with prenatal
alcohol exposure do not present with the characteristic
facial features and therefore need extensive testing with
specialized professionals to determine whether they
meet formal diagnostic criteria for ARND. Moreover,
even if diagnostic clinics are available, wait lists are
typically lengthy and distances to attend such clinics
are often very long. Therefore, there is an urgent need
to develop an FASD screening tool that can expedite a
diagnosis. The findings from the present study offer an
empirical approach for developing such a tool, which
can be used by teachers, social workers, guidance coun-
selors and other frontline professionals to diagnose and
treat children at highest risk.
We presently propose the following FASD screening
tool to be completed by parents or caregivers and used in
such circumstances. This involves a 2-step approach
with the first step identifying behaviours suggestive of
FASD and the second step discriminating children with
FASD from ADHD. This is based on parents’=care-
givers’ responses to the 10 questions shown in Table 4.
If the parent=caregiver answers ‘‘yes’’ to at least 6 of
items 1 to 7, this is suggestive of FASD with 86%
sensitivity and 82% specificity. However, if the child
does not exhibit behaviour consistent with ADHD (i.e.,
the answer is negative for items 2, 6, and 7 or inatten-
tion, impulsivity, hyperactivity), then the child must
receive a score of 3 or more on items 1, 3, 4, and 5 (acts
young, disobedient, lie or cheat, lacks guilt). To rule out
ADHD alone when items 2, 6, or 7 are affirmed, the
child needs to receive a score of 2 or more for items
1, 5, 8, (acts young, lacks guilt, cruelty) or 3 or more for
items 1, 4, 5, 8, 9, 10 (acts young, lies or cheats,
lacks guilt, misbehaves, cruelty, steals at home, steals
To the best of our knowledge, this is the first study to
identify the specific features that characterize the beha-
vioural phenotype in FAS=ARND. In 1998, Streissguth
and colleagues used parent questionnaires to develop a
tool that was able, with high sensitivity, to distinguish
children with FASD from healthy controls (Streissguth
et al, 1998). However, these researchers did not address
the issue of specificity by distinguishing between their
patients and those with other psychopathology, most
notably ADHD. Thus our findings and the proposed
novel tool for screening FASD are applicable in identi-
fying children at highest risk.
Nevertheless, this study is not without limitations.
First, our sample sizes were small and our approach
needs validating in larger groups. Second, our sample
was not young. This is a critical issue when according to
Streissguth and colleagues, a major predictor of fewer
secondary disabilities is the receipt of services before
age 6, because early intervention acts as a buffering
factor (Streissguth et al, 2004). Thus, our approach must
also be replicated on samples of young children with
FASD. A third concern is that we lacked informa-
tion on reported dose and timing of prenatal alcohol
exposure since most cases were not living with a biolo-
gical parent. Similarly, because this information was not
available, our original diagnosis was based soley on how
well the child’s neuropsychological results fit our diag-
nostic criteria, which were developed through findings in
the literature, experience, and empirical testing. Thus,
the present approach needs to be validated in other diag-
nostic clinics that use different approaches for diagnos-
Table 4. Proposed FASD screening tool
1. Does your child act too young for his=her age?
2. Does your child have difficulty concentrating, and can’t pay
attention for long?
3. Is your child disobedient at home?
4. Does your child lie or cheat?
5. Does your child lack guilt after misbehaving?
6. Does your child act impulsively and without thinking?
7. Does your child have difficulty sitting still=is restless=hyperactive?
8. Does your child display acts of cruelty, bullying or meanness
9. Does your child steal from home?
10. Does your child steal outside of home?
Step 1: Identifying behaviour suggestive of FASD
Answering ‘yes’ to at least 6 of items 1–7 is suggestive of FASD
with 86% sensitivity and 82% specificity.
If child does not exhibit behavior consistent with ADHD (i.e.,
parent=caregiver answers ‘no’ to items 2, 6, 7), child must
receive a score of 3 for items 1, 3, 4, and 5.
Step 2: Differentiating FASD from ADHD
Child needs to receive a score of 2 for items 1, 5, 8 or a score of
3 for items 1, 5, 8, 9, 10.
Behavioural phenotype in FASD
ing ARND or on samples followed from pregnancy into Download full-text
later childhood on whom extensive prenatal exposure
data were collected in addition, socioeconomic status
was not controlled for. Fourth, because we did not have
information regarding psychopathology in both biologi-
cal parents, we still do not know whether children’s
behavioural characteristics reflect a genetic predisposi-
tion or are due to prenatal alcohol exposure. Fifth, while
we were successful in differentiating FASD from
ADHD, we do not know whether the features we identi-
fied are unique to FASD or might also be seen in other
pathologic conditions such as conduct disorder and
oppositional defiant disorder. Finally, because our sam-
ple represents clinic-referred children, who are at the
most severe end of the behavioural spectrum, it is impor-
tant that our approach also be evaluated in centers that
deal with less severe cases in order to distinguish those
items that reflect exposure generally from those that
reflect only severe FASD. Further studies stratifying
children by FASD severity may help in identifying those
who need services immediately from those who should
be evaluated further where there is doubt if FASD ser-
vices are really needed. An approach that allows for
triaging patients may help reduce wait-times for assess-
ment as well as costs in assessing every child versus
only those where there is doubt.
Regardless, we are convinced that the present findings
will have important implications for providing appropriate
earlyinterventions to childrenwithlimited accesstobroad
comprehensive diagnostic services and highlighting the
importance of developing alternate screening assessment
strategies for this population. By facilitating earlier diag-
nosis andtreatment ofthe mostsevere cases,thisproposed
screening tool should alleviate some of the burden and
high cost of FASD on society. Since replication of our
approach with very young children is likely to improve
the behavioural disturbances and social cognition deficits
that are most amendable to early intervention and reduce
the risk of later disturbance, it is important that our ap-
Supported in part by grant from the Canadian Institutes for
Health Research. GK holds the Ivey chair in molecular toxi-
cology, University of Western Ontario.
Achenbach T, Rescorla L (2001) Manual for the ASEBA School-Age
Forms and Profiles. University of Vermont, Research Center for
Children Youth & Families, Burlington(VT).
Astley SJ (2004) Diagnostic Guide for Fetal Alcohol Spectrum
Disorders: The 4-Digit Diagnostic Code, 3rd edn. University of
Washington, Seattle, WA.
Biederman J, Monuteaux MC, Kendrick E, Klein KL, Faraone SV
(2005) The CBCL as a screen for psychiatric comorbidity in pae-
diatric patients with ADHD. Arch Dis Child 90: 1010–1015.
Carmichael Olson H, Morse BA, Huffine C (1998) Development of
psychopathology: Fetal Alcohol Syndrome and related conditions.
Semin Clin Neuropsychiatry 3: 262–284.
Centers for Disease Control and Prevention (2004) Fetal Alcohol
Syndrome: Guidelines for referral and diagnosis. Department of
Health and Human Services.
Chudley AE, Contry J, Cook JL, Loock C, Rosales T, LeBlanc N (2005)
Fetal alcohol spectrum disorder: Canadian guidelines for diagnosis.
CMAJ 172 [Suppl 5].
A comparison of children affected by prenatal alcohol exposure and
attention deficit, hyperactivity disorder. Alcohol Clin Exp Res 21:
Greenbaum R (2000) Socioemotional functioning in children with
alcohol related neurodevelopmental disorder (ARND): Profile on
the Child Behavour Checklist (CBCL) Unpublished master’s thesis,
Ontario Institute for Studies in Education, University of Toronto,
Toronto, Ontario, Canada.
Greenbaum R (2004) Socioemotional functioning and language impair-
ment in children with prenatal alcohol exposure: A comparison with
Attention Deficit Hyperactivity Disorder. PhD Thesis, University of
Toronto, Toronto, Ontario, Canada.
Greenbaum R, Nulman I, Rovet J, Koren G (2002) The Toronto
experience in diagnosing alcohol related neurodevelopmental dis-
order (ARND): a unique profile of deficits and assets. Can J Clin
Pharmacol 9: 215–225.
DSM-IVexternalizing disorders with the Child Behaviour Checklist:
a receiver-operating characteristic analysis. J Child Psychol Psychia-
try 45(7): 1299–1307.
Institute of Medicine, National Academy of Science, Stratton K,
epidemiology, prevention, and treatment. National Academy Press,
Jones K, Smith D (1973) Recognition of the fetal alcohol syndrome in
early infancy. Lancet 2: 999–1001.
Nanson JL, Hisock M (1990) Attention deficits in children exposed to
alcohol prenatally. Alcohol Clin Exp Res 14: 656–661.
Sampson PD, Streissguth AP, Bookstein FL, Little RE, Clarren SK,
Dehaene P, Hanson JW, Graham JM Jr (1997) Incidence of fetal
alcohol syndrome and prevalence of alcohol-related neurodevelop-
mental disorder. Teratology 56: 317–326.
Sokol RJ Jr, Delaney-Black V, Nordstrom B (2003) Fetal alcohol
spectrum disorder. JAMA 290: 2996–2999.
Stade B, Ungar W, Stevens B, Koren G (2006) The cost of FAS in
Stratton K, Howe C, Battaglia F (1996) Fetal Alcohol Syndrome:
Diagnosis, epdidemiology, prevention and treatment. Institute of
Medicine, National Academy Press, Washington, DC.
Streissguth AP, Bookstein FL, Barr HM, Press S, Sampson PD (1998) A
fetal alcohol behaviour scale. Alcohol Clin Exp Res 22: 325–333.
JK (2004) Risk factors for adverse life outcomes in fetal alcohol
syndrome and fetal alcohol effects. J Dev Behav Pediatr 25: 228–238.
Correspondence: Gideon Koren, MD, Division of Clinical
Pharmacology, The Hospital for Sick Children, 555 University
Avenue, Toronto, Ontario, M5G 1X8, Canada; e-mail: gkoren@
K. Nash et al: Behavioural phenotype in FASD