Journal of Autism and Developmental Disorders (in press)
Nasal oxytocin for social deficits in childhood autism: A randomized controlled trial.
Mark R Dadds PhD1, Elayne MacDonald MA1, Avril Cauchi B Psych(Hons)1, Katrina
Williams MD3, Florence Levy MD4, & John Brennan MD4.
1. School of Psychology, University of New South Wales, Australia.
2. Neuroscience Research Australia, Australia.
3. University of Melbourne, Murdoch Children’s Research Institute, Australia.
4. School of Psychiatry, University of New South Wales, Australia.
Address for correspondence:
Mark R Dadds
School of Psychology
University of New South Wales
Sydney NSW 2052
Running title: oxytocin and autism
Background: The last two decades have witnessed a surge in research investigating the
application of oxytocin as a method of enhancing social behaviour in humans. Preliminary
evidence suggests oxytocin may have potential as an intervention for autism.
Methods: We evaluated a 5-day ‘live-in’ intervention using a double-blind randomized control
trial. 38 male youths (7–16 years old) with autism spectrum disorders were administered 24IU
or 12IU (depending on weight) intranasal placebo or oxytocin once daily over four consecutive
days. The oxytocin or placebo was administered during parent-child interaction training
sessions. Parent and child behaviours were assessed using parent reports, clinician ratings, and
independent observations, at multiple time points to measure side-effects; social interaction
skills; repetitive behaviours; emotion recognition and diagnostic status.
Results: Compared to placebo, intranasal oxytocin did not significantly improve emotion
recognition, social interaction skills, or general behavioral adjustment in male youths with
autism spectrum disorders.
Conclusions: The results show that the benefits of nasal oxytocin for young individuals with
autism spectrum disorders may be more circumscribed than suggested by previous studies,
and suggest caution in recommended it as an intervention that is broadly effective.
Keywords: Autism, oxytocin, children, randomized controlled trial.
Email for correspondence: email@example.com
Autism Spectrum Disorder (ASD) is characterised by core deficits in social
communication and the presence of repetitive behaviours (Lord et al. 2012). In contrast to
typical child development, children with autism show less interest in other people (Dawson et
al. 2012). While the causes of autism remain unknown, an attentional preference or bias
toward social stimuli is critical to appropriate social development (Adolph et al. 2001).
Impaired attention to critical social stimuli may compromise the developing neural circuitry
that subserves higher social communication domains that are experience-dependent (Dawson
2008; Johnson et al. 2005; Marcus and Nelson 2001), possibly exacerbating or resulting in
the core deficits observed in individuals with ASD.
Oxytocin is a powerful modulator of neural activity that is strongly linked with the
formation of social bonds (Insel 2010). The last two decades have witnessed a surge in
research investigating the application of oxytocin as a method of enhancing social behaviour
in humans. In research involving healthy adults, intranasal oxytocin administration has shown
a range of positive effects such as increasing levels of trust (Kosfeld et al. 2005), gaze to the
eyes (Guastella et al. 2008), and accurate emotion processing (Simplicio et al. 2009;
Ijzendoom and Bakermans-Kranenburg 2012).
There is evidence that oxytocin systems may be disturbed in autism. Common
polymorphisms, and epigenetic methylation of the promoter region of the oxytocin receptor
gene are associated with risk for autism (Campbell et al. 2011; Lerer et a., 2008; Gregory et
al. 2009). There is evidence that children with autism demonstrate lower plasma oxytocin
levels (Modahl et al. 1998), and increased levels of oxytocin precursor peptides compared to
controls (Green et al. 2001); although these results have not been found reliably in young
people with ASD (Miller et al., 2013). Collectively, a growing but tentative body of evidence
is accumulating that reduced oxytocinergic function may be a contributing factor to an
endophenotype underlying social deficits in ASD.
Considering this evidence for the importance of oxytocin in social behaviour, the core
deficits in ASD, and the possible reduced oxytocinergic function in ASD individuals, several
researchers have proposed a role for synthetic oxytocin as a pharmacological treatment (Insel
2010; MacDonald and MacDonald 2010). So far, a handful of studies have examined the
effects of nasal and intravenous synthetic oxytocin on individuals with ASD, and all have
reported positive results, including a reduction in repetitive behaviours (Hollander et al.
2003), and increases in social memory (Hollander et al. 2007) and emotion processing
(Guastella et al. 2010; Andari et al. 2010). While these studies augur for a role of oxytocin in
the treatment of autism; they are limited to small sample sizes (largest n = 16; Guastella et al.,
2010), single dose administrations of oxytocin aiming to produce specific behavioural and/or
cognitive effects, rather than broad ‘treatment’ studies. One recent study (Anagostou et al.,
2012) examined repeated administrations of intranasal oxytocin or placebo with n = 19 adults
with autism (twice daily over 6 weeks). They found no improvements on primary measures
of social function and repetitive behaviour; however the oxytocin group performed better on
emotion recognition tasks and a quality of life measure. With regard to safety in young
people, Tachibana et al. (2013) showed no adverse effects of long-term oxytocin use in a
small sample of early adolescent boys with ASD.
Thus, there are preliminary data to suggest that nasal oxytocin may help with social
functioning in autism; however, larger trials are needed. This study tested the role of oxytocin
in potentiating the performance, development and generalisation of interpersonal social skills
in young children with ASD. We conducted a randomised controlled trial of four consecutive
daily administrations of oxytocin during parent-child interactions to address the following
questions: 1) does oxytocin improve social communication skills, (i.e. eye contact, warmth,
verbal content)? ; 2) does oxytocin reduce repetitive behaviours? ; 3) does oxytocin improve
emotion recognition and 4) does oxytocin bring about generalised improvements beyond
Participants were N = 54 male children recruited through Royal Far West, Sydney
Australia, between January 2010 and January 2012 (aged 7-16 years; M=11.23, SD=2.6). All
met DSM-IV criteria for Autistic Disorder, Asperger’s Disorder or PDD-NOS (American
Psychiatric Association 2000) using a multi-stage diagnostic procedure described below. In-
clusion criteria required a diagnosis of ASD, having English as the first language, and IQ=80
or above on a standardized intelligence test. Exclusion criteria were female gender, allergy to
preservatives, major comorbid illness such as epilepsy or heart conditions. We limited this
initial study to males as ASD is more common in males and there is evidence that oxytocin
may impact differently on females (De Vries 2008).
The participant sample had a high level of comorbid disorders that are commonly
associated with ASD (Simonoff et al. 2008). Twenty participants had comorbid Attention
Deficit Hyperactivity Disorder; 13 had a diagnosis of Oppositional defiant disorder, and 6
had internalizing anxiety disorders. Seventeen participants were stabilized for over 8 weeks
on psychotropic medication (Concerta n=8; Risperidol 2; Catapress 1; Ritalin 3;
Dexamphetamine 3). Figure 1 shows recruitment and retention through the study and table 1
lists clinical variables and socio-demographics.
Child diagnoses was made by a specialist child psychiatrist using convergent
information from existing referral diagnoses, the DISCAP-ASD diagnostic clinical interview,
and observational assessment using the Childhood Autism Rating Scale (Schopler et al.
1988), the High Functioning Autism Spectrum Screening Questionnaire (ASSQ) (Ehlers et al.
1999), the OSU Autism Rating Scale-DSM IV (OSU Research Unit 2005a) and the OSU
Autism Global Impression Scale (OSU Research Unit 2005b). To check the reliability, 20%
of diagnoses were reviewed by a multi-disciplinary team who were ‘blind’ to the primary
clinician’s diagnosis; inter-rater reliability kappa = 0.82 (p < 0.001).
IQ scores were gathered using the Wechsler Intelligence Scale (Wechsler 2005). All
IQ tests where conducted by a trained clinician or had been done within the last 2 years.
Participants also completed a medical review with a Pediatrician. Ethical approval was
provided by the University of New South Wales Ethics Committee (09133) and The Research
Institute at The Children’s Hospital Westmead (09/CHW/79). This trial was registered with
the Australian New Zealand Clinical Trials Registry (ACTRN12609000784213) and the
Australian Government Therapeutic Goods Administration (TGA) 2009/009858(163).
Participants were assessed twice prior to treatment: Initial (3-6 months before
treatment) and Pretreatment (immediately before treatment), three times during treatment,
immediately Post-treatment, and at 3-month follow-up. The treatment was delivered over five
consecutive days and involved four key components; (1) nasal administration; (2) parent-child
interaction training with the treating psychologist; (3) family interaction task (30 minutes); (4)
outcome measures . Participant’s and their parents stayed in Royal Far West accommodation
for the duration of the treatment week. The attended only for the treatment study and had no
other appointments during their stay. Following drug treatment randomization (oxytocin OT or
placebo PL), participants were further randomized into two treatment orders: Group 1 (PL =
10; OT 10) completed the family interaction task at 9.30am each day (Tues-Thurs) and the
parent-child interaction training at 11.30am each day (Mon-Thurs). Group 2 (PL=9; OT=9)
completed the family interaction task at 11.30am each day (Tues-Thurs) and the parent-child
interaction training at 9.30am each day (Mon-Thurs). Each group received the nasal
administration at 11am (Mon/Wed) and 9am (Tues/Thurs). Table 1 supplementary information
shows the complete treatment design. Regardless of order of delivery, each participant received
two sessions where spray administrations (oxytocin or placebo) were prior to parent-child
interaction training and two prior to the family interaction tasks; all observational data were
collected for the latter.
Preparation: Sprays were prepared by the University of New South Wales pharmacist. Each
puff per nostril contained 6 international units (IU) of oxytocin, mannitol, glycerine, methyl
parraben, propyl, parraben and purified water (placebo contained all ingredients except the
active oxytocin and mannitol). Each participant received a total of four doses (one daily over
four consecutive days) 30-45 minutes before experimental procedures. Participants weighing
40+kg received 24IU (n=21, OT=10, PL=11), delivered as two puffs to each nostril, the
standard dose usually given to adults (MacDonald et al. 2011) and participants under 40kgs
received half the adult dose at 12IU (n=17, OT=9 PL=8), delivered as one puff to each nostril.
Participants were instructed to abstain from alcohol and caffeine on the day of drug
administration and food and drink (except water) two hours before drug administration.
Administration: Oxytocin administration was consistent with the recent recommendations of
Guastella et al. (2012). The experimenter primed each nasal spray by pumping the spray until
a fine mist was observed, thus removing any displaced air present in the tube. Then all
participants were given clear instructions to hold one nostril closed and to breath in through
the nose immediately following nasal administration to reduce any gravitational effects (this
was practiced a couple of times prior to administration). The experimenter asked participants
to slightly tilt head back and placed the spray approximately 50-100mm inside the nostril at a
30-45º angle. The nasal spray was then administered followed by three ‘sniff-like’
inhalations modeled by the experimenter. This process was completed once in each nostril for
12IU participants and twice in each nostril for 24IU participants, any issues with the
procedures were recorded. There were two participants who received 3 doses instead of 4 due
to feeling unwell on the day (cold/flu like symptoms). Both of these participants were
receiving 12IU, one was placebo and the other oxytocin. There were only two individual
puffs (6IU in each puff) where it appeared that a reduced amount of solution left the bottle
(different participants, both receiving OT). All participants were comfortable to receive the
nasal spray, although a handful reported a strange smell and/or that they could feel some of it
go down their throat.
Parent-Child Interaction Training
The key objective of this treatment study was to evaluate the effectiveness of nasal
oxytocin for reducing social deficits in autism. A critical issue concerns the context in which
the oxytocin should be administered. We evaluated the effects of administration in a
therapeutic context that: 1) provided rich opportunities for positive affiliative behaviour, and
2) had some empirical support as being useful for promoting positive social behaviour for
ASD. Thus, our Parent-Child Interaction Training was selected based on current knowledge
of evidence-based psychological treatments for autism and specifically to complement the
use of nasal oxytocin and consisted of two key intervention components, 1) teaching emotion
recognition and 2) teaching key social interaction skills.
The emotion recognition training used the Mindreading (MR) program developed by
Simon Baron-Cohen (Baron-Cohen 2007). The second part of this training program involved
using positive video feedback to enhance the way in which both parent and child interact with
each other. This approach is based on “Video Interactive Guidance” (VIG) (Kennedy 2011
for a comprehensive review of VIG) and involves using short video clips demonstrating the
clients use of successful communication skill, such as, eye contact, positive body language
and responding to others. Research has widely supported the use of video as a way of
improving social and interpersonal skills in both adults and children with autism (Reichow
and Volkmar 2010; Bellini and Akullian 2007; Kroeger et al. 2007).
The experimenter was a Child Psychologist with a Masters Degree and a Post
Graduate Diploma in Autism. Furthermore the experimenter participated in formal training in
VIG, which involved a 2 day initial training course, 7 hours of VIG supervision and 0.5
accreditation day. The experimenter conducted all of the intervention sessions with the parent
and child. As we combined both emotion training and the use of video we created a
comprehensive therapist manual to accompany the training program, this also promoted
consistency across participants. During the final session a home task workbook was given to
the child and parent which was to be completed prior to the three month follow-up visit.
Family Interaction Task
Parent-child dyads completed the Family Interaction Task at each assessment time-
point. The specific tasks included: Free Play (10 minutes), Emotion Talk (10 minutes) and I-
Love-You Task (2 minutes). In Free Play the parent and child are given a range of games/toys
and asked to play together as they like for 10 minutes. In Emotion Talk the parent and child
are asked to discuss a happy and sad time that they have shared together. This task aims to
explore the child and parent’s generally conversation skills, and more specifically their ability
to discuss with each other emotion based conversational topics. The I-Love-You Task
explores how the child responds to having his parent express positive emotion to the child
(Dadds et al. 2012). The primary objective of the Family Interaction Task was to gather
observational data. All of the Family Interaction Tasks were video-taped and later coded by a
two trained video coders. The experimenter was not present during the Family Interaction
Task and the observational data was extracted for the initial assessment, pre-treatment, time
point 1-3, post-treatment and 3 month-post. The Family Observation Schedule-ASD (FOS-
ASD) (MacDonald and Dadds 2010) was used as the coding instrument for scoring the
parent-child interactions during the Family Interaction Tasks. This schedule was adapted
from the FOS-6th Edition (Pasalich and Dadds 2009) for use with families of children with
ASD. The theoretical underpinnings of this coding instrument are embedded in the
behavioural principles of social learning theory (Patterson 1982); attachment theory (Bowlby
et al. 1992) and inter-subjectivity (Trevarthen and Aitken 2001). The codes reflect the
behavioral and affective aspects of parent-child interactions which are required for successful
social interactions and attachment formations. The FOS-ASD incorporates an amalgamated
procedure for coding family interaction, bringing together the social learning micro-coding
approach; tallying the frequency of behaviours occurring within discrete time-intervals, with
the attachment macro-coding approach; globally scoring behaviours along a continuum.
There were two video coders, both blind to treatment conditions. Both coders were trained
using the coding manual (MacDonald and Dadds 2010) and practice videos. Twenty percent
of videos were coded independently by the two raters to check for inter-rater reliability; ICC
= 0.801, p<0.001 (95% CI = .49 - .92).
Parent and child behaviours were assessed at multiple time points for: side-effects,
social interaction skills, repetitive behaviours, emotion recognition and generalised effects
(diagnostic change). Time points were Initial assessment; Pre-treatment; Time 1 (T1); Time 2
(T2); Time 3 (T3); Post-treatment; three-month post treatment. All observational data were
collected during the family assessment tasks. Two of the assessment points (T1 & T3) were
immediately (30-45 mins) after the child received oxytocin (or placebo). Time point 2 (T2) was
used to assess the generalizability of oxytocin (or placebo) mid-treatment (after two parent-
child interaction sessions and oxytocin or placebo administrations), but not after receiving
oxytocin (or placebo) immediately beforehand.
Side-effects: Side-effects were monitored throughout the study. Parents completed a detailed
physiological checklist at each time point (designed in consultation with a paediatrician and
child psychiatrist and with reference to available information on oxytocin safety (Novartis,
2009). Participant blood pressure and heart rate was also recorded throughout the treatment,
once at time point 1 and 2, before and after each nasal spray administration, and at time point 7
and 8. The parent and child were asked if they thought the child has received oxytocin or
placebo each day following administration. The experimenter also recorded their own thoughts
as to what each participant had received.
Social interaction: Social interaction was measured through parental questionnaires, video
micro-coding and global coding of observations. Parental questionnaires included; the Social
Skills Rating Scale (SSRS: Gresham and Elliot 1990). Video analysis of the family observation
task took place at all 7 time points. A five-point likert scale (0-4) was used with a higher score
indicating that more of the behaviour was occurring. Each video was coded for a global rating
of social interaction which consisted of talk; warmth; responsiveness and eye contact. Each
video was also micro-coded for positive body language, verbal content and asking questions
over each 30 seconds of video time. At the end of each 30 second interval the video was
stopped and parent and child were coded for eye contact (present=1 or absent=0). The coding
manual is available from the first author.
Repetitive Behaviours: Repetitive behaviours were measured through micro-analysis of the
family observation videos at all time points. The parents also completed the Social Reciprocity
Scale which rated autistic mannerisms (Constantino and Gruber 2005) (both given at initial,
pretreatment, posttreatment and 3 month posttreatment).
Emotion recognition: All participants completed the UNSW Facial Emotion task (Dadds et al.
2004). In this task the participant views sets of happy, sad, angry, fearful, disgusted, and neutral
faces on a computer monitor (1-sec duration) and is asked to identify the emotion. This
measure has established reliability and validity for measuring fear recognition in children
(Dadds et al. 2008). Overall accuracy scores for each emotion were obtained.
Generalised effects/diagnostic status: At initial contact and then at three month post treatment,
participants diagnosis was re-assessed using the OSU, CARS and the DISCAP-ASD as
Design and statistical analysis
The study was a randomized-controlled trial comparing oxytocin (OT) with placebo
(PL) with participants and their parents, investigator team, outcome assessors, family
interaction coders and data analyzers blind to treatment group. All statistical analyses were
conducted using SPSS 20.0 (SPSS Inc, Chicago, IL, USA). A preliminary one-way ANOVA
was conducted on demographic variables and pre-treatment diagnostic variables to ensure
that the groups had been randomly assigned – there were no differences between groups.
Drop-out after randomization was negligible (n=4) and equally distributed across groups;
thus, missing data at the case level was trivial and treatment effects were evaluated using
repeated measures ANOVAs. Missing values analysis of observational data indicated there
was small amounts of missing-at-random data distributed across the data file; these were
video-data and assessments lost due to equipment failure or participants missing a session. To
correct for this Rueben’s method of multiple imputation was used.
The treatment order effects (supplementary table 1) were not significant, F(1,21) =
1.190, p=.288, thus all further analysis was conducted on oxytocin versus placebo without
reference to treatment order. All observational data were collected during the family
assessment tasks. Two of the assessment points (T1 & T3) were immediately (30-45 mins)
after the child received oxytocin (or placebo). Time point 2 (T2) was used to assess the
generalizability of oxytocin (or placebo) mid-treatment (after two parent-child interaction
sessions and oxytocin or placebo administrations), but not after receiving oxytocin (or
placebo) immediately beforehand. Significant interaction effects were dismantled using
Tukey’s HSD tests.
Demographics and clinical sample
Figure 1 shows recruitment and retention through the study. Thirty-five participants
completed time points 1-7; n=three participants failed to return for the three month follow-up.
There were no significant baseline differences between groups in demographics or clinical
variables (Table 1). There was a high level of co-morbid diagnosis which was equally
distributed between treatment groups. Use of psychotropic medication was also equally
distributed between groups, χ2(33)=0.79, p =0.37.
Safety, side effects and subjective awareness
Participants reported minimal side-effects throughout the study. There was a
significant main effect for time, (F(3.62, 115.87) = 7.96, p<0.0), both groups reported
decreased side effects from pretreatment to post. There were no significant group effect (F(1,
32) = 1.48, p=0.23), and no time x group interaction (F(3.62, 115.87) = 0.59, p = 0.76).
Similarly, we found no significant main effects or interaction for group and time on systolic
and diastolic blood pressure, and heart rate. Neither children nor parents were able to guess
15 Download full-text
whether they had received oxytocin or placebo (p>0.1); however, guesses by the
experimenter were borderline significant; that is, she was slightly above chance in guessing
whether a participant had received oxytocin or placebo, χ2(1) = 4.0, p =0.046.
Social interaction skills
Micro-analysis of Verbal and Nonverbal communication;
Child eye contact: There was a significant main effect for time with child eye contact
increasing from Initial to Time 3, (F(6, 216) = 4.70, p <0.001). There was no main effect for
group, (F(1,36)=0.30 p= 0.59), or time x group, (F(6 ,216)=1.58, p=0.16) (Figure 2).
Parent eye contact: We found a significant main effect for time, (F(6, 216) = 3.05, p = 0.01),
with parent eye contact increasing significantly from Pre-treatment to Time 3. There was no
main effect for group, (F(1,36) = 0.87, p = 0.35) or time x group, (F(6, 216) = 1.34, p= 0.24)
Positive non-verbal behaviours: There was a significant main effect for time on child positive
non-verbal behaviours from Pretreatment-Posttreatment, (F(6, 216)=7.51, p<0.001); however,
there was no main effect for group, (F(1,36)=0.58, p= 0.45) or time x group, (F(6, 216) =
1.71, p=0.12) (Figure 2).
Verbal Content: Micro analysis of child verbal content, asking questions and giving
information, found no significant main effect for time, (F(6, 216)=0.64, p=0 .70), group (F(1,
36)=0.89, p= 0.35) or time x group, (F(6, 216) = 0.82, p=0.55) (Figure 2).
Global ratings of social interaction
The video analysis of global ratings found a significant main effect on the quality of child
social interaction skills over time, (F(4.49, 161.72) = 3.105, p=0.01), child social interaction