UC San Diego
UC San Diego Previously Published Works
Title
Non-ASD outcomes at 36 months in siblings at familial risk for autism spectrum disorder
(ASD): A baby siblings research consortium (BSRC) study.
Permalink
https://escholarship.org/uc/item/6kk319nr
Journal
Autism research : official journal of the International Society for Autism Research, 10(1)
ISSN
1939-3792
Authors
Charman, Tony
Young, Gregory S
Brian, Jessica
et al.
Publication Date
2017
DOI
10.1002/aur.1669
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
RESEARCH ARTICLE
Non-ASD Outcomes at 36 Months in Siblings at Familial Risk for
Autism Spectrum Disorder (ASD): A Baby Siblings Research
Consortium (BSRC) Study
Tony Charman, Gregory S. Young, Jessica Brian, Alice Carter, Leslie J. Carver, Katarzyna Chawarska,
Suzanne Curtin, Karen Dobkins, Mayada Elsabbagh, Stelios Georgiades, Irva Hertz-Picciotto,
Ted Hutman, Jana M. Iverson, Emily J. Jones, Rebecca Landa, Suzanne Macari, Daniel S. Messinger,
Charles A. Ne lson, Sally Ozonoff, Celine Saulnier, Wendy L. Stone, Helen Tager-Flusberg,
Sara Jane Webb, Nurit Yirmiya, and Lonnie Zwaigenbaum
LAY ABSTRACT
This study characterized developmental outcomes of a large sample of siblings at familial high-risk of autism spec-
trum disorder (ASD), who themselves did not have ASD (n 5 859), and low-risk controls with no family history of
ASD (n 5 473). We characterized outcomes at age 3 years using a developmental assessment of language and learning
and an observational measure of ASD symptoms and, where available, parent interviews about ASD behaviors and
adaptive functioning. Around one-in-ten high-risk siblings had mild-to-moderate levels of developmental delay, a
rate significantly higher than the low-risk controls. The groups did not differ in the proportion of toddlers with mild-
to-moderate language delay. High-risk siblings were also more likely to have higher levels of observer-rated and
parent-reported levels of ASD symptoms and lower adaptive functioning. Males were more likely to show higher lev-
els of ASD symptoms and lower levels of developmental ability and adaptive behavior than females across most meas-
ures. Lower maternal education was associated with lower developmental and adaptive behavior outcomes. We
discuss these findings as evidence for early emerging characteristics related to the “broader autism phenotype” previ-
ously described in older family members of individuals with ASD. There is a need for ongoing clinical monitoring of
high-risk siblings who do not show clear signs of ASD by age 3 years, as well as continued follow-up into school age
to determine their developmental and behavioral outcomes.
SCIENTIFIC ABSTRACT
We characterized developmental outcomes of a large sample of siblings at familial high-risk of autism spectrum disor-
der (ASD), who themselves did not have ASD (n 5 859), and low-risk controls with no family history of ASD (n 5 473).
We report outcomes at age 3 years using the Mullen Scales of Early Learning, the Autism Diagnostic Observation
Schedule (ADOS), the Autism Diagnostic Interview—Revised (ADI-R) and adaptive functioning on the Vineland Adap-
tive Behavior Scales. Around 11% of high-risk siblings had mild-to-moderate levels of developmental delay, a rate
higher than the low-risk controls. The groups did not differ in the proportion of toddlers with mild-to-moderate lan-
guage delay. Thirty percent of high-risk siblings had elevated scores on the ADOS, double the rate seen in the low-
risk controls. High-risk siblings also had higher parent reported levels of ASD symptoms on the ADI-R and lower
adaptive functioning on the Vineland. Males were more likely to show higher levels of ASD symptoms and lower lev-
els of developmental ability and adaptive behavior than females across most measures but not mild-to-moderate lan-
guage delay. Lower maternal education was associated with lower developmental and adaptive behavior outcomes.
These findings are evidence for early emerging characteristics related to the “broader autism phenotype” (BAP) previ-
ously described in older family members of individuals with ASD. There is a need for ongoing clinical monitoring of
high-risk siblings who do not have an ASD by age 3 years, as well as continued follow-up into school age to deter-
mine their developmental and behavioral outcomes. Autism Res 2016, 00: 000–000.
V
C
2016 International Society
for Autism Research, Wiley Periodicals, Inc.
Keywords: autism spectrum disorder; broader autism phenotype; developmental outcomes; high risk siblings; adaptive
functioning
From the King’s Coll ege London (T.C.); University of California, Davis (G.S.Y., I.H.P., S.O.); University of Toronto (J.B.); University of Massachusetts,
Boston (A.C.); University of California, San Diego (L.J.C., S.M.); Yale University School of Medicine (K.C.); University of Calgary (S.C.); McGill Uni-
versity (M.E.); McMaster University (S.G.); University of California, Los Angeles (T.H.); University of Pittsburgh (J.M.I.); Birkbeck College London
(E.J.J.); Kennedy Krieger Institute and John Hopkins School of Medicine (R.L. ); University of Miami (D.S.M.); Harvard Medical School (C.A.N.); Har-
vard Graduate School of Education (C.A.N.); Boston Children’s Hospital (C.A.N.); Emory University School of Medicine (C.S., S.J.W.); University of
Washington (W.L.S.); Boston University (H.T.F.); Seattle Children’s Research Institute (S.J.W.); Hebrew University of Jerusalem (N.Y.); University of
Alberta (L.Z.)
Received December 06, 2015; accepted for publication June 13, 2016
Address for correspondence and reprints: Tony Charman, Department of Psychology PO 77, Institute of Psychiatry, Psychology & Neuroscience,
King’s College London, De Crespigny Park, London SE5 8AF, UK. E-mail: tony.charman@kcl.ac.uk
Published online 00 Month 2016 in Wiley Online Library (wileyonlinelibrary.com)
DOI: 10.1002/aur.1669
V
C
2016 International Society for Autism Research, Wiley Periodicals, Inc.
INSAR
Autism Research 00: 00–00, 2016 1
Introduction
Research on infant siblings at familial high-risk (HR) of
autism spectrum disorder (ASD) has established that
close to 20% of HR siblings have ASD themselves by
the age of 36 months. Using pooled data from collabo-
rating sites in the Baby Siblings Research Consortium
(BSRC), recurrence was 18.7% from a sample of
n 5 664 HR siblings [Ozonoff et al., 2011] and 19.5% in
an expanded cohort of n 5 1241 HR siblings [Messinger
et al., 2015]. A previous report from the BSRC on HR
siblings (n 5 507) who did not have an ASD outcome at
36 months used latent class analysis to subgroup these
children [Messinger et al., 2013]. Groups were formed
based on scores on a symptom measure (Autism Diag-
nostic Observation Schedule (ADOS); Lord et al., 2000)
and a standardized developmental assessment (Mullen
Scales of Early Learning (MSEL); Mullen, 1995). Twenty-
one percent of non-ASD HR siblings were classified in
groups with higher ASD severity scores and/or lower
levels of developmental function, a profile found in
only 7% of low-risk (LR) controls without a family his-
tory of ASD [Messinger et al., 2013].
Understanding more about non-ASD outcomes in HR
siblings at an early age would allow us to study the ear-
ly emergence of the broader autism phenotype (BAP)—
subclinical traits or characteristics that are present at an
elevated rate in families containing individuals with
autism [Bolton et al., 1994; Folstein & Rutter, 1977;
Pickles et al., 2000; Piven et al., 1997]. The term “BAP”
has been used in different ways in the literature, some-
times including only subclinical features closely aligned
to the core diagnostic features of ASD (e.g., social com-
munication and pragmatic language difficulties, behav-
ioral rigidity) and other times referring to a broader
range of characteristics that are elevated in family
members and associated with, but not “core” to, the
definition of ASD (e.g., co-occurring psychiatric disor-
ders, intellectual disability) (see Sucksmith, Roth, &
Hoekstra, 2011; for a review).
Groups using the prospective HR sibling design have
reported non-ASD developmental outcomes (see
Szatmari et al., 2016; for a review). Elevated rates of
sub-clinical ASD symptoms (characterized as the BAP),
symptoms of emergent attention deficit hyperactivity
disorder (ADHD), and lower language and developmen-
tal abilities were found in around one quarter of HR sib-
lings who did not have ASD at 36 months of age in a
sample from two BSRC sites [Ozonoff et al., 2014].
Another report from an overlapping cohort found ele-
vated rates of pragmatic language difficulties in HR sib-
lings who do not have ASD at 36 months; in most cases
these children did not have more general language
impairments [Miller et al., 2015].
The current study utilizes an expanded BSRC HR sam-
ple (relative to Messinger et al., 2013) and reports on
the outcomes at 36 months of age in HR infants who
do not have ASD (n 5 859) compared to LR controls,
also without ASD (n 5 473). In contrast to the statisti-
cally derived classification reported by Messinger et al.
2013, we adopt a more clinical framework by reporting
outcomes in terms of children who have mild-to-
moderate global developmental and/or language delays
and those who exhibit elevated scores on the ADOS but
who were not categorized as having ASD. The current
approach therefore allows us to indicate non-ASD devel-
opmental outcomes for individual children, which was
not possible with the latent class approach taken by
Messinger et al. 2013. Where available, additional infor-
mation on autism symptoms (Autism Diagnostic Inter-
view–Revised (ADI-R); Lord, Rutter, & Lecouteur, 1994)
and adaptive functioning (Vineland Adaptive Behavior
Scales (Vineland); Sparrow, Cicchetti, & Balla, 1984,
2005) is also used to characterize outcomes. Together,
this information allows us to report rates of non-ASD
developmental difficulties (mild-to-moderate develop-
mental and language delay) as well as sub-clinical levels
of ASD symptoms and adaptive behavior relevant to the
emergent BAP in these HR non-ASD siblings.
Methods
Participants
In line with previous BSRC reports [Chawarska et al.,
2014; Messinger et al., 2013; Ozonoff et al., 2011], ASD
case definition was a consensus best estimate (CBE)
diagnosis of ASD (using DSM-IV [American Psychiatric
Association (APA), 2000], DSM-5 [APA, 2013] or ICD-10
[World Health Organisation (WHO), 1993] criteria) and
scoring at or above the ASD threshold on the ADOS
(calibrated severity score (CSS) 4; Gotham, Pickles, &
Lord, 2009). Of the 2099 infants/toddlers included in
this dataset, 620 were low-risk (LR) infants (no first
degree relative with ASD), of whom 3 (0.48%) met ASD
criteria at age 3 years, and 1479 were high-risk (HR)
infants (at least 1 older sibling with ASD), of whom 275
(18.59%) met ASD criteria. The aim of the article is to
describe 36 month outcomes in HR siblings who do not
have ASD, so the 278 cases meeting CBE ASD criteria
(from both the HR and LR groups) were removed from
the sample. This left 1204 HR siblings and 617 LR
infants. The primary outcome measures are the ADOS-
CSS and the MSEL, so children missing either or both
were excluded (2 sites did not provide ADOS data and a
different 2 sites did not provide MSEL data; from other
sites ADOS and MSEL data were incomplete), leaving
n 5 859 HR siblings and n 5 473 LR controls (total
2 Charman et al./Non-ASD outcomes at 36 months in high-risk siblings
INSAR
n 5 1332) from 9 sites. Vineland (n 5 895) and ADI-R
(n 5 600) data were available on a sub-set of sample.
Measures
The Mullen Scales of Early Learning (MSEL; Mullen, 1995)
is a standardized developmental assessment for children
aged between birth and 68 months. It yields a global
development quotient, the Early Learning Composite
(ELC), with a mean of 100 and a standard deviation of
15. We report T-scores (mean of 50, SD 10) averaged
across the two verbal (Expressive language, Receptive
language) and two non-verbal (Fine motor, Visual
reception) subscales.
The Autism Diagnostic Observation Schedule (ADOS;
Lord et al., 2000) is a play-based, observer-rated assess-
ment of ASD symptoms. Different modules are used
depending on the child’s language ability and, at the
36 month assessment, 141 children completed Module
1 (no words or single words only) and 1191 children
completed Module 2 (phrase speech). The ADOS-CSS
ranges from 1 to 10, with the threshold for an ASD
diagnosis being 4 or greater [Gotham et al., 2009].
The Autism Diagnostic Interview—Revised (ADI-R; Lord
et al., 1994) is an informant-based, examiner-rated
interview of ASD symptoms. It yields domain scores
covering social (“reciprocal social interaction”), com-
munication (“communication and language”), and
repetitive (“restricted and repetitive, stereotyped inter-
ests and behaviors”) symptoms.
The Vineland Adaptive Behavior Scales [Sparrow et al.,
1984, 2005] is an informant interview assessing every-
day adaptive functioning. It yields Socialization, Com-
munication, Daily Living Skills, and Motor domain
scores and an overall Adaptive Behavior Composite
score (ABC), all having a mean of 100 and a standard
deviation of 15. Approximately half the sample had
Vineland scores from the first edition (Sparrow et al.,
1984; n 5 423) and half from the second edition
(Sparrow et al., 2005, n 5 472).
Characterising developmental outcomes
We characterized atypical developmental outcomes in
terms of: (1) elevated ASD symptom expression as
assessed by the ADOS and ADI-R; and (2) below average
general developmental and language abilities as
assessed by the MSEL and adaptive function as assessed
by the Vineland. We first defined mild-to-moderate
Developmental Delay as a Mullen ELC >1 SD below the
mean (i.e., below 85) and then mild-to-moderate Lan-
guage Delay as expressive language (EL) and/or recep-
tive language (RL) > 1SD below the mean (i.e., T-score
below 40), so the two subgroups were mutually exclu-
sive. We defined “elevated” ASD symptoms using an
ADOS CSS threshold of 3 (where three is one point
below the ASD diagnostic threshold) to include sub-
threshold levels of ASD behaviors [see Chawarska et al.,
2014]. For the ADI-R we used “sub-clinical-threshold”
cut points of 8 for the Social, 6 for the Communica-
tion, and 2 for the Stereotyped, Repetitive and Rigid
Behaviour (RRB) domains, respectively. For the Vine-
land we defined mild-to-moderate adaptive behavior
delay as a standardized score >1 SD below the mean
(i.e., below 85) on each domain score and/or the ABC.
Statistical analysis
Proportions of HR siblings and LR controls falling into
each categorical outcome group were analysed by mul-
tinomial logistic regression (for MSEL-defined nominal
outcomes) and logistic regression (for the remaining
dichotomous outcomes) with relative risk ratios (RRR),
odds ratios (OR), and 95% confidence intervals being
reported, respectively. Background variables on which
the HR and LR groups were significantly different were
entered first and retained in the models when they
were significantly associated to outcomes (see below).
Risk group (HR vs. LR) and sex (male vs. female) were
then entered into the models. The interaction between
sex and risk group was then entered and retained if sig-
nificant. Finally, BSRC site was entered into the models.
We report the proportion of male and female HR and
LR children falling into each outcome group in the
main Tables but we also present the continuous scores
on the measures in Supporting Information Tables S1–
S4 to aid comparison with previous literature.
Results
Sample characteristics are shown in Table 1. The LR
and HR groups were comparable on background varia-
bles with the exception of age-first-seen, which was
higher in the HR siblings compared to the LR controls
Table 1. Sample Characteristics
Low Risk
(N 5 473)
High Risk
(N 5 859)
Sex (% Male) 53.07% 50.06%
Non-Caucasian (%)
a
18.85% 19.84%
Maternal Education
(% college or higher)
b
82.18%*** 72.75%
Age first seen
(months; mean (SD))
7.00 (3.67)** 7.76 (4.43)
Age at outcome
(months; mean (SD))
c
37.40 (2.32) 37.53 (2.44)
a
n 5 733;
b
n 5 1194;
c
n 5 1271;
**P < 0.01;
***P < 0.001.
INSAR
Charman et al./Non-ASD outcomes at 36 months in high-risk siblings 3
(F(1, 1330) 5 10.25, P < 0.01), and maternal education,
which was higher in the LR controls than in the HR sib-
lings (v
2
(1, N 5 1194) 5 13.78, P < 0.001). Both variables
were entered first into the statistical models and
retained where significant.
Rates of Developmental Delay and Language Delay
Table 2 shows the rates of the HR and LR children fall-
ing into the MSEL-defined mild-to-moderate Develop-
mental Delay and Language Delay groups. A
multinomial logistic regression indicated that both
(lower) maternal education and (higher) age-first-seen
were significantly associated with Developmental Delay
and Language Delay outcomes (likelihood ratio (LR) v
2
(2, n 5 1194) 5 49.25, P < 0.001). The relative risk ratio
(RRR) of being in the Developmental Delay group (vs.
the No Developmental/No Language Delay group) was
0.59 (95% CIs: 0.48, 0.73, P < 0.001) across the 5-point
maternal education scale and 0.60 (0.47, 0.78,
P < 0.001) of being in the Language Delay group. The
RRR of being in the Developmental Delay group was
1.08 (1.02, 1.13, P < 0.01) for each month of age and
1.09 (1.03, 1.16, P < 0.01) of being in the Language
Delay group.
1
With both maternal education and age-
first-seen retained (LR v
2
(6, n 5 1194) 5 83.43,
P < 0.001), risk group and sex were added to the model
that remained significant (LR v
2
(8, n 5 1194) 5 99.02,
P < 0.001). The RRR of being in the Developmental
Delay group for the HR vs. the LR group was 2.84 (1.62,
5.01, P < 0.001) and 4.01 (2.41, 6.68, P < 0.001) for
males vs. females. Lower maternal education continued
to be associated with a higher risk of being in the
Developmental Delay group (P < 0.001) as did higher
age-first-seen (P < 0.05). Neither risk group (P 5 0.45)
nor sex (P 5 0.11) was associated with being in the Lan-
guage Delay group, but lower maternal education
(P < 0.001) and higher age-first-seen (P < 0.01) were. The
interaction of sex and risk group was not significant
with being in the Developmental Delay (P 5 0.58) or
Language Delay (P 5 0.38) groups. BSRC site was
entered last and was not associated with being in either
the Developmental Delay group (P 5 0.49) or the Lan-
guage Delay group (P 5 0.21).
Elevated levels of ASD Symptoms
The proportion of children with an elevated ADOS-CSS
(3) was higher in the HR group (n 5 250/859, 29.10%)
compared to the LR group (n 5 80/393, 16.91%)—see
Table 3. A logistic regression (Likelihood ratio (LR) v
2
(2, N 5 11 194) 5 2.08, P 5 0.35) indicated that neither
maternal education (P 5 0.77) nor age-first-seen
(P 5 0.16) was associated with an elevated ADOS CSS
and both were dropped from the model. When risk
group and sex were entered, the overall model was sig-
nificant ((LR) v
2
(2, N 5 1332) 5 34.42, P < 0.001). There
was a main effect of risk group for being in the elevated
ADOS CSS group for HR vs. LR status ((OR) 5 2.05 (1.54,
2.72), P < 0.001) and a main effect of sex for males vs.
females ((OR) 5 1.47 (1.14, 1.90), P < 0.01). The interac-
tion of sex by risk group was not significant (P 5 0.95).
There was also a main effect of BSRC site (P < 0.05).
Table 4 shows elevated ADI-R scores by risk group
and sex. For the ADI-R Social domain, a logistic regres-
sion ((LR) v
2
(2, N 5 512) 5 4.59, P 5 0.10) indicated
that neither maternal education (P 5 0.07) nor age-first-
seen (P 5 0.18) was associated with an elevated score.
When risk group and sex were entered, the overall
model was significant ((LR) v
2
(2, N 5 600) 5 12.88,
P < 0.01). There was a main effect of risk group for HR
vs. LR ((OR) 5 4.37 (1.30, 14.67), P < 0.05) and a main
Table 2. Mullen-Defined Outcomes (Developmental Delay and Language Delay) by Risk Group and Sex
Low Risk non-ASD High Risk non-ASD
Total
(N 5 473)
Females
(N 5 222)
Males
(N 5 251)
Total
(N 5 859)
Females
(N 5 429)
Males
(N 5 430)
No Delay 433 (91.54%) 206 (92.97%) 227 (90.44%) 709 (82.54%) 386 (89.98%) 323 (75.12%)
Developmental Delay (ELC < 85) 16 (3.38%) 4 (1.80%) 12 (4.78%) 91 (10.59%) 19 (4.43%) 72 (16.74%)
Language Delay (T-score < 40) 24 (5.07%) 12 (5.41%) 12 (4.78%) 59 (6.87%) 24 (5.59%) 35 (8.14%)
Table 3. Elevated ADOS Scores (Total CSS 3-and-Above) by Risk Group and Sex
Low Risk non-ASD High Risk non-ASD
Total
(N 5 473)
Females
(N 5 222)
Males
(N 5 251)
Total
(N 5 859)
Females
(N 5 429)
Males
(N 5 430)
ADOS CSS < 3 393 (83.09%) 191 (86.04%) 202 (80.48%) 609 (70.90%) 321 (74.83%) 288 (66.96%)
ADOS CSS > 53 80 (16.91%) 31 (13.96%) 49 (19.52%) 250 (29.10%) 108 (25.17%) 142 (33.02%)
1
When infants first seen above 12 months of age were excluded, age-
first-seen was no longer associated with mild-to-moderate Developmen-
tal Delay or mild-to-moderate Language Delay.
4 Charman et al./Non-ASD outcomes at 36 months in high-risk siblings
INSAR