Introduction

Research in Autism Spectrum Disorders 6 (2012) 546–555

Contents lists available at ScienceDirect

Research in Autism Spectrum Disorders

Jo u rn al h om ep ag e: h t tp : / /ees .e lsev ier . co m /RASD/d efau l t .as p

Psychological adjustment and sibling relationships in siblings of children with Autism Spectrum Disorders: Environmental stressors and the Broad Autism Phenotype

Michael A. Petalas a,*, Richard P. Hastings a, Susie Nash a, Louise M. Hall a,b, Helen Joannidi a,b, Alan Dowey a,b

a School of Psychology, Bangor University, Bangor, Gwynedd LL57 2AS, UK b Wrexham Child Health Centre, Betsi Cadwaladr University Health Board, UK

A R T I C L E I N F O

Article history:

Received 27 March 2011

Received in revised form 19 July 2011

Accepted 19 July 2011

Keywords:

Psychological adjustment

Autism

Behaviour problems

Broad Autism Phenotype

Sibling relationships

Parents

A B S T R A C T

Research with siblings of children with Autism Spectrum Disorders (ASD) suggests that

they may be at increased risk for behavioural and emotional problems and relatively poor

sibling relationships. This study investigated a diathesis-stress model, whereby the

presence of Broad Autism Phenotype features in the typically developing siblings might

interact with family-environmental risk variables to predict sibling functioning (5–17

years of age) of children with an Autism Spectrum Disorder (ASD), their child with an ASD,

and their own psychological well-being. Sibling adjustment was associated with the

extent of behaviour problems in the child with an ASD and with the extent of the sibling’s

Broad Autism Phenotype (BAP) features. Sibling relationships were more negative when

the child with an ASD had more behaviour problems and when there was evidence of

critical expressed emotion in the family environment. Siblings with more BAP features,

who had brothers/sisters with an ASD and a greater number of behaviour problems, had

more behaviour problems themselves. Siblings with more BAP features who had parents

with mental health problems reported more sibling relationship conflict.

� 2011 Published by Elsevier Ltd.

1. Introduction

Siblings growing up with a brother or sister with autism may experience emotional and behavioural difficulties (Petalas, Hastings, Nash, Lloyd, & Dowey, 2009), and difficulties in their relationships (Kaminsky & Dewey, 2001) with their brother or sister with an Autism Spectrum Disorder (ASD). However, other siblings avoid such adverse adjustment outcomes (Benson & Karlof, 2008; Hastings, 2007), enjoying warm relationships with their brother or sister with autism (Fisman et al., 1996) and potentially deriving benefit through their unique circumstances and experiences (Petalas et al., 2009). Previous studies reviewing the literature on the psychological well-being and sibling relationships of siblings of children with an ASD have highlighted methodological problems, such as the use of different comparison groups and outcome measures across studies, and a lack of consensus in findings (e.g., Bauminger & Yirmiya, 2001).

The variability observed in sibling adjustment and relationships led to a shift away from a unidirectional model focusing on the adverse impact of the child with an ASD on the adjustment of the typically developing (TD) sibling and the quality of the sibling relationship. Researchers have increasingly adopted ecological and family systems perspectives when examining

* Corresponding author. Tel.: +44(0)7901536201.

E-mail address: [email protected] (M.A. Petalas).

1750-9467/$ – see front matter � 2011 Published by Elsevier Ltd.

doi:10.1016/j.rasd.2011.07.015

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555 547

sibling adjustment (e.g., Fisman et al., 1996). The role of factors such as parental distress (Fisman et al., 1996), siblings’ perception of parental partiality (Wolf, Fisman, Ellison, & Freeman, 1998), and family climate (Benson & Karlof, 2008) have been examined.

Although research conducted from a family-systems perspective improves on the simplistic model of focusing solely on the impact of the child with autism, there has been neglect of the putative impact of the genetic component that is characteristic of ASD. Studies comparing relatives of individuals with an ASD to TD controls have described the presence of subtle features that closely resemble autistic characteristics, referred to as the Broad Autism Phenotype (BAP). In siblings, the BAP has been described in terms of difficulties in language, cognition, socialization, and emotional experience (e.g., Bolton, Pickles, Murphy, & Rutter, 1998).

Bauminger and Yirmiya (2001) urged researchers to pursue a more integrated biological-genetic and environmental- family approach to study sibling adjustment. They proposed a diathesis-stress model, highlighting the interplay between genetic liability (e.g., BAP-related impairments in problem-solving ability or social-emotional functioning), and various environmental stressors (e.g., parental availability for the TD sibling). Orsmond and Seltzer (2009), applied this diathesis- stress model to research with 57 adolescent siblings of children with an ASD. They found that siblings presenting with BAP characteristics (sub-threshold ASD symptoms) had elevated symptoms of depression and anxiety, but only in the presence of increased stressful life events. Depressive symptoms in siblings were most pronounced when they had more sub-threshold autism characteristics themselves, and had experienced a higher number of life events in the past year. Sibling anxiety was highest when they had more sub-threshold autism characteristics themselves and a mother with high levels of depression symptoms.

In the present study, our main aim was to explore the interactions between genetic liability (i.e., the BAP) and environmental stressors in predicting the adjustment of siblings of individuals with an ASD and the quality of their sibling relationships. Orsmond, Kuo, and Seltzer (2009, p. 76) called for the need to examine the effect of the BAP on both sibling adjustment and sibling relationship outcomes. The sibling relationship between the TD sibling and the child with an ASD plays a pertinent role in the TD sibling’s development and subsequent adjustment. In TD sibling dyads, the sibling relationship and sibling adjustment outcomes are interlinked. For example, siblings who reported less warmth, intimacy, and a lack of friendly behaviour in their sibling relationships showed increased internalizing and externalizing behaviour problems (Dunn, Slomkowski, Beardsall, & Rende, 1994). Increased conflict between siblings has been identified as a risk factor for behaviour problems in young children and adolescents (Slomkowski, Rende, Conger, Simons, & Conger, 2001). In contrast, increased positivity, in the form of affection in sibling relationships, is strongly associated with more favourable child adjustment outcomes (Pike, Coldwell, & Dunn, 2005).

Given the potential significance of sibling relationships, we explored whether the diathesis-stress model might help to explain both adjustment and sibling relationship outcomes. We used a psychometrically robust measure of the BAP (Auyeung, Baron-Cohen, Wheelwright, & Allison, 2007; Baron-Cohen, Hoekstra, Knickmeyer, & Wheelwright, 2006), and selected environmental stressors based on previous research findings. Previous research has shown the adjustment of siblings to be related to the behaviour problems of their brother or sister with an ASD (Hastings, 2007), parental psychological distress (Fisman et al., 1996), lower socioeconomic status (Petalas et al., 2009), which is considered to be an environmental risk variable in many areas of child and adolescent psychopathology (Conger et al., 2002), and negative family climate (Benson & Karlof, 2008). Benson and Karlof (2008) assessed family climate by adding together scores across three constructs: family connectedness (extent of parent and child participation in activities jointly), parent agreement (extent of parental agreement on key family issues e.g., parenting practices), and child marital impact (extent of parental perception of positive impact of child with an ASD on marital relationship).

In this research we measured family climate via expressed emotion (EE). Expressed emotion is a construct used to measure familial emotional climate between close family members and their ‘‘ill’’ relative. In the present research, we extended the concept of familial EE by defining a critical family environment as one where the primary carer’s relationship with their child with an ASD or their TD child included relatively high criticism (e.g., use of highly critical evaluative phrases such as ‘‘I hate it’’ when speaking about their child’s behaviour). Within families with TD children, EE has been explored as a family level risk variable such that familial EE (either or both parents coded as high in overall EE) was found to differentiate children with depression from controls (Asarnow, Tompson, Woo, & Cantwell, 2001). In a sample of 149 mothers residing with their adolescent or adult child with autism, high expressed emotion was related to increased levels of behaviour problems and more severe symptoms of autism over 18 months (Greenberg, Seltzer, Hong, & Orsmond, 2006). High parental criticism in particular has been found to be associated with more behaviour problems in adolescents and adults with autism (Orsmond, Seltzer, Greenberg, & Krauss, 2006) as well as in children with other developmental disabilities (Hastings & Lloyd, 2007).

High EE, especially criticism, within any central relationship within the family will be reflected in an emotionally charged atmosphere and so may affect all members of the family system (e.g., McCarthy, Lau, Valeri, & Weisz, 2004). This tentative hypothesis was examined in the present research. We explored the main and interacting effects of the BAP and environmental stressor variables. We hypothesized that siblings with a more pronounced BAP alongside higher levels of environmental stress would be those with the most psychological problems themselves and less positive sibling relationships as measured by the sibling relationship questionnaire (SRQ; Buhrmester & Furman, 1990). This would include lower scores on the warmth/closeness factor (e.g., reduced affection), and higher scores on the conflict factor (e.g., increased quarrelling and antagonism).

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555548

2. Method

2.1. Participants

One hundred and sixty-six families participated in the research. The parental primary caregivers were between 26 and 64 years of age, with a mean age of 42.25 years (SD = 5.17). Eight were male, 158 were female; there were 156 mothers, eight fathers, one foster parent, and one adoptive parent. In terms of their ethnicity, all described themselves as White British, or White Irish, and one stated Mixed White and Black Caribbean. Seventy-nine (47.6%) of the parents had a university postgraduate or undergraduate degree, and 87 (52.4%) did not have a university degree. One hundred and forty-one (84.9%) parents were married and living with a spouse/partner, and 25 (15.1%) were divorced, separated, single or widowed and not living with a partner. The number of children currently living in the family home ranged from two to six (mode = 2; mean = 2.47, SD = .69). Eighty-three (50%) parents reported a family income of above £35,000 (approximately 50–55,000 US dollars) and 81 (48.8%) reported an income below this level (two parents (1.2%) did not provide any income information). Overall, 60% of parental primary caregivers also worked outside of the home. Of these, 25 worked full-time and 80 worked part-time.

The children with an ASD were 137 boys and 29 girls. Their ages ranged between five and 17 years with a mean age of 10.46 years (SD = 2.73). Based on parental report, 100 children had a diagnosis of autism and comorbid intellectual disability, 65 children had an Asperger’s syndrome diagnosis, and one child had a diagnosis of pervasive developmental disorder not otherwise specified. Fifty-two children had been diagnosed by a clinical psychologist, 62 had been diagnosed by a paediatrician, 30 children had received their diagnosis through a multidisciplinary team, 12 children had been diagnosed by a psychiatrist, and for 10 cases information on who provided the diagnosis was not provided. Twenty-four children attended mainstream school, 86 attended mainstream school with support, 34 children attended a special school, 12 children attended a specialist autism unit in mainstream school, and 10 children attended other educational services.

The main study sample was 166 TD siblings of children with an ASD. The sibling participants were chosen by the primary caregivers, who were asked to select the TD sibling closest in age to the child with an ASD. According to parental report, these siblings did not have a disability or psychiatric diagnosis. There were 84 brothers and 82 sisters of siblings of children with autism. Their ages ranged from five to 17 years (mean age = 10.49, SD = 3.40). Eighty-three siblings were younger than the child with an ASD, 73 were older, and eight were twins (missing data N = 2). Eighty-five siblings were the same gender as the child with autism, 81 were different. There were 149 biological siblings of children with an ASD, and three half biological siblings (missing data N = 14). Sixty-six siblings attended the same school as their brother or sister with an ASD, while 100 siblings attended a different school.

2.2. Measures

Six measures were used in addition to a demographics questionnaire, which was used to collect basic background information (see Section 2.1) about the parents, the child with an ASD, and the TD sibling closest in age to the child with an ASD. As part of the demographics questionnaire, parents were asked to provide information on the total family income as a measure of socio-economic circumstances.

2.2.1. Strengths and Difficulties Questionnaire (SDQ)

Parents completed the SDQ (Goodman, 1997) as a measure of emotional and behavioural adjustment for the child with an ASD and the TD sibling. The SDQ is a 25 item screening measure with four problem domains, assessing emotional problems (e.g., ‘‘often unhappy, downhearted or tearful’’), conduct problems (e.g., ‘‘often has temper tantrums’’), hyperactivity (e.g., ‘‘easily distracted’’), and peer relationship problems (e.g., ‘‘has at least one good friend’’), as well as a positive behaviour domain (prosocial behaviour, e.g., ‘‘Considerate of other people’s feelings’’). A total difficulties score is derived by summing the total ratings of the four problem domains. The SDQ is a frequently employed and well-validated measure of child behavioural adjustment, against clinical interview criteria (Pike et al., 2005). It has been shown to be as effective as the Child Behaviour Checklist (Achenbach, 1991) and the Rutter Scales (Elander & Rutter, 1996) in identifying clinically significant levels of behavioural disturbance in children (Goodman, 1997; Goodman & Scott, 1999). Previous research with siblings of children with developmental disabilities and children with autism and intellectual disability, using the SDQ, suggests that good levels of reliability are maintained in these populations (Hastings, 2007). Cronbach’s alpha coefficients for the current total sample of siblings of children with an ASD and TD siblings respectively on the SDQ subscales were as follows: Children with an ASD .77 for prosocial behaviour, and .78 for total difficulties; TD siblings .80 for prosocial behaviour, and .88 for total difficulties.

2.2.2. Autism Spectrum Quotient (AQ)

The AQ is a self-report measure that was originally developed to quantify autistic traits in adults. The AQ has been shown to have reasonable face validity (adults with an ASD scored significantly higher on the AQ compared to matched controls), and reasonable construct validity (each of the five domains measured (social, communication, imagination, attention to detail, and attention to switching) show moderate to high a coefficients) (Baron-Cohen, Wheelwright, Skinner, Martin, & Clubley, 2001). The AQ was subsequently adapted into parent/caregiver-report versions for adolescents aged 12+ years

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555 549

(Baron-Cohen et al., 2006) and children aged 4–11 years (AQ-Child; Auyeung et al., 2007), which were the versions used in this study. The two versions are directly comparable. The adolescent version is a 50-item questionnaire, comprised of five domains: social skills (e.g., ‘‘S/he prefers to do things with others rather than on her/his own’’), attention to switching (e.g., ‘‘S/he prefers to do things the same way over and over again’’), attention to detail (e.g., ‘‘S/he often notices small sounds when others do not’’), communication (e.g., ‘‘S/he enjoys social chit-chat’’), and imagination (e.g., ‘‘If s/he tries to imagine something, s/he finds it very easy to create a picture in her/his mind’’), each assessed through 10 questions. The AQ adolescent has been shown to have reasonable face validity; a mean of 89.3% percent of adolescents with an ASD scored above a critical minimum of 30 compared to none of the controls; and reasonable construct validity (each of the five domains show high a coefficients). The adolescent AQ has moderate to good levels of internal consistency (Cronbach’s a coefficients for all five domains range from .60 to .90, and for the AQ total score .79) and high test–retest reliability (r = .92) (Baron-Cohen et al., 2006). Cronbach’s alpha coefficients for the current sample of siblings of children with an ASD for the total adolescent AQ score was .82.

The AQ-Child is also a 50-item questionnaire, comprised of the same five domains as the adolescent version, each assessed through 10 questions. The AQ-Child shows high specificity (0.95) and sensitivity (0.95), and 95% of individuals with an ASD diagnosis scored at or above the cut-off score (76) used. The AQ-Child has good levels of internal consistency (Cronbach’s a coefficients for all five domains range from .80 to .90, and for the AQ-Child total score .97) and high test–retest reliability (r = .85) (Auyeung et al., 2007). Cronbach’s alpha coefficient for the current sample of siblings of children with an ASD for the AQ-Child total score was .78.

2.2.3. Sibling Relationship Questionnaire – revised (SRQ brief version)

The SRQ brief version (Buhrmester & Furman, 1990), is a 39-item questionnaire, measuring 16 dimensions of sibling relationship contributing to four scales: warmth/closeness (e.g., ‘‘How much do ______ and this sibling go places and do things together?’’), relative status/power (e.g., ‘‘How much does _______ tell this sibling what to do?’’), conflict (e.g., ‘‘How much do ______ and this sibling disagree and quarrel with each other?’’), and rivalry (e.g., ‘‘Who usually gets treated better by mother, _______ or this sibling?’’). On the parent-report version of the SRQ, which was used in this study, the parent/ caregiver is asked to rate how well a particular characteristic describes the relationship between two siblings on a five-point likert scale ranging from hardly ever true to extremely much. We were granted permission by the SRQ authors to adapt the wording of the questionnaire for use with a British-English speaking population, and some minor wording alterations have been made (a copy of the version used in this study is available from the first author upon request). Research has shown adequate construct validity of the SRQ in a sample of 428 Dutch adolescents (e.g., warmth/closeness was negatively correlated with internalizing (r = �.16) and externalizing (r = �.23) behaviours, whereas conflict was positively correlated with internalizing (r = .13) and externalizing (r = .16) behaviours) (Derkman, Scholte, Van der Veld, & Engels, 2010).

The SRQ brief version has been used previously in research with siblings of children with an ASD (Fisman et al., 1996; Kaminsky & Dewey, 2001). Warmth/closeness, relative status/power, and conflict are scored by summing rated items within these domains. The rivalry score is derived by averaging items for maternal partiality and paternal partiality. For participants with one parent only, the missing values were replaced with a case-wise weighted mean for completed items. The SRQ has been shown to have acceptable test–retest reliability, low correlations with social desirability, and adequate construct validity (Buhrmester & Furman, 1990). Internal consistency coefficients, for the SRQ range from .71 to .81 for children aged 8, 11, 14, and 17 years, and reported test–retest reliability is .71 (Buhrmester & Furman, 1990). Cronbach’s alpha coefficients for the current sample of sibling dyads were: warmth/closeness = .91, relative status/power = .63, conflict = .87, and rivalry = .90. Given the poor reliability of the relative status/power scale, this was excluded from the statistical analyses.

2.2.4. Five Minute Speech Sample (FMSS)

The FMSS (Magana et al., 1986) was used as a measure of the parents’ critical expressed emotion (EE) toward the child with autism and the TD sibling (i.e., to assess emotional climate within the family). Speech samples were collected during telephone interviews with the parent. The interviews were counterbalanced with respect to the sequence of the child (ASD/ TD) that the interviewee reported on. The interviewer instructed the parent to talk uninterrupted for five minutes about their thoughts, feelings and relationship with the target child focusing on the preceding six months. Speech samples were recorded and subsequently coded using the FMSS scoring manual (Magana et al., 1986). Two raters were involved in the EE coding process; they were both trained in the administration and coding of EE using the FMSS. Inter-rater reliability checks were carried out, during which one of the raters was always blind to the status of the target child (i.e. child with an ASD vs. TD sibling). Being truly blind was not always possible as the content of the primary caregiver’s conversation at times informed the rater if the child had an ASD. Categorical scores for critical climate were created by coding primary caregivers as high or borderline versus low for criticism scores. Combining scores for borderline and high criticism has been used in previous research (e.g., Yan, Hammen, Cohen, Daley, & Henry, 2004). Negative family climate was categorized as present when parents expressed high or borderline criticism toward one or both children. A rating of high criticism applied if the parent met one or more of the following criteria: they made a negative initial statement about the target child, they described having a negative relationship with the target child or made one or more critical comments about the target child. A borderline coding applied if a primary caregiver made a comment about the target child that implied dissatisfaction but was not strong enough to be considered a criticism. Low criticism was scored when parents did not meet high or borderline ratings.

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555550

Research with relatives of individuals diagnosed with schizophrenia using the FMSS showed good concurrent validity with the Camberwell Family Interview (CFI), which is considered the gold standard for measuring EE (Magana et al., 1986). In the field of intellectual disabilities, high EE has been shown to possess predictive validity for behaviour problems of the child with an intellectual disability. High EE as measured by the FMSS in particular has been positively associated with parental reports of caregiving burden, maternal stress, maternal pessimism, internalizing and externalizing child behaviour problems and asocial behaviour (Hastings & Lloyd, 2007). Previous research using the FMSS has shown good inter-rater reliability with mothers of individuals with autism (Percentage Agreement Index = 83.3%, kappa = .67) (Greenberg et al., 2006). In the present research, code–recode reliability (25 speech samples coded and recoded by the same person four weeks apart) for criticism was high (Percentage Agreement Index = 92%, Cohen’s kappa = .84), as was inter-rater reliability (20 speech samples coded by different raters) for criticism (Percentage Agreement Index = 95%, Cohen’s kappa = .90).

2.2.5. Hospital Anxiety and Depression Scale (HADS)

The HADS (Zigmond & Snaith, 1983) was used as a measure of parental mental health. This is a reliable and valid measure that has been used with both residential as well as community samples including parents of children with autism (e.g., Hastings et al., 2005). The HADS demonstrates excellent sensitivity and specificity in the ranges of 0.70–0.90, and excellent clinical case-finding abilities. The concurrent validity of the HADS has been demonstrated against other measures of depression and anxiety (e.g., general health questionnaire) with correlations between .60 and .80 (Bjelland, Dahl, Haug, & Necklemann, 2002). The HADS is a 14-item questionnaire that can be divided into two subscales, one assessing for depression and one for anxiety, each consisting of seven items. Scores from 8 to 10 on each scale represent a mild clinical disorder, scores from 11 to 21 indicate moderate clinical disorder, and scores of 16 and above are indicative of a severe clinical disorder (Zigmond & Snaith, 1983). In the present study, each item was rated on a 0–3 scale. A cut-off score of 11 was employed. A dichotomous variable ‘‘parent mental health problems’’ was created based on whether or not the parent scored 11+ on either the anxiety or the depression subscale or both.

2.3. Procedure

Following receipt of ethical approval for the research, 1000 invitations were given to a national autism charity to distribute amongst families who had previously attended one of their introductory support programmes. Three hundred and five families, who met the inclusion criteria (presence of a child with autism between the ages of five to 17 years, and a primary caregiver present in the home), expressed an interest in the research by returning the invitation with their personal contact details. Questionnaire packs were sent out to the families, and a member of the research team telephoned the families for an interview with the primary caregiver. Of the 305 families who were mailed questionnaires, responses were received from 215. If questionnaires were not returned within two weeks of being sent out, a personally addressed reminder letter was mailed to the home. The present research focuses on 166 families where there was also a TD sibling between the ages of 5 and 17 years, with no known disabilities or diagnosed psychiatric disorder.

3. Results

3.1. Exploratory analyses

All of the demographic variables described in the participants section were tested for associations with siblings’ total difficulties and prosocial behaviour on the SDQ, and warmth/closeness, conflict, and rivalry on the SRQ. Correlation analyses revealed a negative association between the length of time since diagnosis of the child with an ASD and conflict (r(162) = �.39, p < .001), and rivalry scores (r(162) = �.19, p = .015) on the SRQ. Siblings whose parents had university level education had lower scores on SDQ total difficulties (t(164) = 3.35, p = .001), and also had a lower conflict score on the SRQ (t(163) = 2.91, p = .004). Finally, siblings older than the child with an ASD had lower conflict scores on the SRQ than those who were younger (t(153) = 2.87, p = .005). No other relationships with demographic variables were statistically significant and so are not reported here.

3.2. Main analyses

The main analyses focused on five regression models. In each model, control variables were introduced (i.e., the demographic variables with significant relationships with sibling dependent variables) but otherwise predictor variables were identical: sibling AQ score, child with autism SDQ total difficulties, whether the parent scored at a clinically significant level for anxiety or depression symptoms or both (51% of primary caregivers were coded as presenting with moderate or severe levels of anxiety and/or depression), critical family climate (43% of primary caregivers were coded as high/borderline in EE criticism toward either the child with an ASD, the TD sibling, or both children; 13 primary caregivers expressed high or borderline criticism toward the TD sibling only, 45 primary caregivers expressed high or borderline criticism toward the child with an ASD only, and 14 primary caregivers expressed high or borderline criticism toward both the TD sibling and the child with an ASD), and total family income. Four interaction terms were then generated by multiplying the sibling AQ z- score with the z-scores of each of the four environmental stressors (child with autism SDQ total difficulties; clinically

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555 551

significant levels for anxiety or depression symptoms in parent, critical family climate, and family income). These interaction terms were included in all regression models to test the diathesis-stress predictions.

3.2.1. Regression analyses – main effects

The results of the regression analyses are summarized in Tables 1 and 2. A number of main effect relationships were observed. Total difficulties on the SDQ of the child with an ASD was a

significant independent positive predictor of sibling SDQ total difficulties and conflict and rivalry in the sibling relationship, and also a negative predictor of warmth in the sibling relationship. Sibling AQ score was a positive predictor of sibling SDQ total difficulties and a negative predictor of sibling prosocial behaviour. Critical family climate predicted conflict in the sibling relationship. Finally, total family income was a marginal negative predictor of rivalry in the sibling relationship. Relative age of the siblings and length of time since diagnosis of the child with an ASD also emerged as significant independent predictors of conflict in the sibling relationship. Specifically, siblings older than the child with an ASD and whose brother or sister had been longer diagnosed had less parent-reported conflict in the sibling relationship.

3.2.2. Regression analyses – interacting effects

Two interaction terms also emerged as statistically significant predictors in the regression models: the interaction between the child’s with an ASD SDQ total difficulties and sibling AQ score for sibling SDQ total difficulties, and between parent mental health problems and sibling AQ for conflict in the sibling relationship. To explore the nature of these interaction effects, the procedures outlined by Aiken and West (1991) were followed. For sibling SDQ total difficulties, values were entered into the regression equation for SDQ total difficulties of the child with an ASD at high (1 SD above the mean) and low (1 SD below the mean) levels crossed with three levels of sibling AQ scores (high, low, and medium – at the mean value). The results of this analysis are displayed in Fig. 1. Sibling SDQ total difficulties were highest when they had more autism characteristics themselves and when environmental stress was high (i.e., their sibling with an ASD had a higher SDQ total difficulties score). An identical procedure was used to explore the parent mental health-sibling AQ interaction, and the results of this analysis are displayed in Fig. 2. Parents reported more conflict in the sibling relationship when both the BAP risk (sibling AQ score) and environmental stressor (presence of parental mental health problem) were high/present.

Table 1

Regression analysis of sibling total difficulties and sibling prosocial behaviour.

Predictor Sibling SDQ total difficultiesa Sibling prosocial behaviorb

b p b p

Parent level of education �.146 .089 – –

Sibling AQ .366 <.001 �.345 <.001

Child with an ASD SDQ total difficulties .220 .007 �.029 .742

Critical family environment .015 .846 .002 .985

Parent mental health problem �.118 .135 .045 .598

Total family income �.041 .625 .103 .226

Sibling AQ � child with an ASD SDQ total difficulties .170 .045 �.131 .156

Sibling AQ � critical family environment .056 .459 .045 .586

Sibling AQ � parent mental health problem .025 .772 .006 .951

Sibling AQ � total family income �.098 .197 .152 .069 a Model R = .59, R2 = .35, F (10,126) = 6.71, p < .001. b Model R = .45, R2 = .20, F (10,129) = 3.58, p = .001.

Table 2

Regression analysis of sibling relationship factors (warmth/closeness, conflict, rivalry).

Predictor Warmth/closenessa Conflictb Rivalryc

b p b p b p

Sibling relative age – – �.189 .010 – –

Child with an ASD time since diagnosis – – �.295 <.001 �.122 .141

Parent level of education – – .019 .822 �.106 .275

Sibling AQ .017 .850 .105 .174 .089 .317

Child with an ASD SDQ total difficulties �.273 .004 .282 <.001 .204 .025

Critical family environment �.064 .472 .233 .002 .072 .402

Parent mental health problem .006 .945 �.008 .921 �.219 .014

Total family income �.035 .695 �.163 .047 .138 .142

Sibling AQ � child with an ASD SDQ total difficulties �.009 .931 .009 910 �.027 .776

Sibling AQ � critical family environment .039 .654 �.024 .738 �.039 .642

Sibling AQ � parent mental health problem .059 .560 .173 .042 �.097 .319

Sibling AQ � total family income .006 .950 �.047 .527 .050 .558 a Model R = .31, R2 = .09, F (9,126) = 1.46, p = .171. b Model R = .64, R2 = .41, F (12,121) = 7.06, p < .001. c Model R = .45, R2 = .20, F (11,123) = 2.84, p = .002.

5

7.5

10

12.5

15

17.5

Low SDQ Total Difficulties of Child wi th an

ASD

High SDQ Tota l Difficulties of Child with an

ASD Si

bl in

g SD

Q T

ot al

D iff

ic ul

tie s

Low Sibling AQ Medium Sibling AQ High Sibling AQ

Fig. 1. Interpretation of the interaction between child with an ASD total difficulties and sibling AQ in predicting sibling total difficulties.

2.5

3

3.5

Low Parent Mental Health

Problems

High Parent Mental Health

Problem s

C on

fli ct

in S

ib lin

g R

el at

io ns

hi p

Low Sibling AQ Score Medium Sibling AQ Score High Sibling AQ Score

Fig. 2. Interpretation of the interaction between parent mental health problem and sibling AQ in predicting conflict in the sibling relationship.

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555552

4. Discussion

The results of the present research both replicate and extend previous research findings. First, the behaviour problems of the child with an ASD were a predictor of increased behaviour problems in siblings (cf. Hastings, 2007). Behaviour problems of the child with an ASD also predicted less reported warmth and more conflict and rivalry in the sibling relationship. This finding is consistent with recent research showing that behaviour problems negatively impact the sibling relationship, as siblings may be less willing to interact with their brother or sister with an ASD who has behavioural difficulties (Orsmond et al., 2009). Second, demographic factors were predictive of sibling conflict. Specifically, the longer the time since the child with an ASD had received a diagnosis the less sibling conflict was reported (cf. Ferrari, 1984), and when siblings were younger than the child with an ASD more conflict was reported between siblings (cf. Petalas et al., 2009).

Original findings resulted from our exploration of critical family environments (as assessed with the Five Minute Speech Sample), and sibling BAP. When mothers were rated as critical toward either their child with an ASD, their sibling, or toward both children, more conflict was reported between the siblings. These findings extend family research by Orsmond et al. (2006), who assessed levels of maternal expressed emotion in 202 adolescents and adults with autism. Their results showed high levels of maternal criticism to be associated with more behaviour problems in adolescents and adults with autism.

In the present research, the BAP had both main and interacting effects. Thus, those siblings with a more pronounced BAP also had increased behaviour problems and reduced prosocial behaviour scores. While the diathesis stress model might not typically be applied to positive (prosocial) behaviour, a lack of prosocial skills might be considered as an area of psychological deficit (cf. Petalas et al., 2009). Sibling BAP was not predictive of sibling relationship quality in any domain. The main effect in the prediction of sibling behaviour problems was moderated by the extent of the behaviour problems of the child with an ASD such that siblings with higher BAP scores and a brother or sister with more behaviour problems were most at risk of behaviour problems themselves. There was also an interaction between sibling BAP and the presence of parental health problems such that siblings with higher BAP scores and parents with likely mental health problems were reported to have more conflict in their relationships with their siblings. The interaction effect is further notable in the context of previous research showing mothers’ adjustment to be affected by the behaviour problems of the child with autism (Hastings et al., 2005). One may speculate, for example, that the behaviour problems of the child with an ASD might mediate the relationship between parental mental health and conflict in the sibling relationship, which in turn is moderated by sibling BAP traits. The finding that, in the context of siblings’ low BAP characteristics, higher parental mental health problems were associated with

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555 553

lower levels of conflict in the sibling relationship was unexpected. One hypothesis is that children with fewer BAP characteristics may be better able to empathize with their parents with mental health difficulties, and elect to engage in less conflictual relationships and/or take on more helping and nurturing roles with their siblings with an ASD. These hypotheses would require attention in future research should this finding be replicated.

The interaction effects between BAP and environmental risk variables (behaviour problems of the child with an ASD, and parent mental health) were consistent with the predictions from a diathesis-stress model (Bauminger & Yirmiya, 2001), as well as previous research showing higher risk of adjustment difficulties in siblings of children with an ASD who have increased environmental (higher number of negative life events) and putative genetic (higher BAP characteristics) susceptibility (Orsmond & Seltzer, 2009). Thus, this model is worthy of more attention in research on sibling adjustment and sibling relationships in families of children with an ASD. In the present research, the exact processes by which risk variables contribute to negative sibling adjustment, decreased prosocial behaviour, and difficulties in the sibling relationship are not clear. To establish causal links between risk factors and adjustment outcomes future research needs to explore how changes in a given risk variable temporally precede changes in the outcome variable using longitudinal designs (Hastings, 2007).

The BAP, as assessed by the AQ measures, may not represent a purely biological risk factor. In particular, it could be argued that at least some of the characteristics assessed by the AQ might be learned from the sibling with an ASD (especially if that sibling is older) and/or might be exacerbated by psychosocial stressors. Related this point is the issue of potential overlap between BAP symptoms and emotional and behavioural symptoms, both in terms of measurement as well as in the constructs themselves, which may be contributing to the main-effect associations between them. The interaction findings provide some assurance that measurement/construct overlap is not a substantial issue. However, in the absence of robust bio-markers for autism, autistic traits have to be measured at the behavioural level. It is additionally worth mentioning that the findings noted here were generated using BAP as a broad dimensional construct and a small percentage of undiagnosed siblings are likely to present with significant BAP traits.

A further limitation of the current study is that parents provided data on their own mental health as well as reporting on the child with an ASD and their sibling. Thus, source variance is likely to be a problem and multiple sources of data including potentially secondary caregiver/fathers’ reports and siblings’ self-reports need to be explored in future research. Future research exploring the contribution of the BAP should also consider the influence of parental BAP features. There might have been some siblings who scored high on the AQ who also had parents with significant BAP features, which may present an additional risk variable.

At least five additional limitations of the research need to be considered. First, the diagnosis of an ASD was conveyed solely through parental report. Second, the level of intellectual and adaptive functioning of the children with an ASD was not measured and might have contributed to the findings. Following on from these two methodological issues, the severity and frequency of problem behaviours exhibited by the child with an ASD may vary as a function of the type of ASD diagnosis. In addition, the presence (or absence) of a comorbid intellectual disability is closely related to diagnosis. These factors may have important effects on the relationship between siblings where one child has an ASD and future research should separate these variables in the analyses. Within our study, we explored whether autism versus aspergers as a dichotomous variable was related to differences for the sibling relationship or sibling well-being on the SDQ. There were no differences at univariate level when we compared sibling relationship and well-being when the sibling had a brother or sister with autism vs. Asperger syndrome, with the exception that sibling relationship conflict differed between these two group. However, adding the autism versus aspergers variable in the regression model for conflict did not alter the pattern of results and this diagnosis dichotomous variable also was not a significant independent predictor of sibling conflict in the regression model. Thus, our results remain when we account for this diagnostic difference. However, in using this exploratory strategy we were not able to separate the children with autism and an associated intellectual disability from the children with autism without an intellectual disability, as we did not possess data on autism and intellectual disability comorbidity. Future research should endeavor to gather such data and to examine potential differences. Third, eight of the siblings were twins of their brother or sister with an ASD. Although this is a small number within the overall sample, these siblings may well have had a stronger underlying BAP leading to potential bias in the results. To explore the potential impact of these siblings’ data, we repeated the main regression analyses excluding data from these eight siblings and found the same pattern of results. A similar argument could be made for including half-siblings. Again, in repeating the main regression analyses excluding data from the three half-siblings did not alter the pattern of results. Thus, although twin versus non-twin effects, and full versus half biological siblings, might be important to explore in future research the presence of a small number of twins and half- siblings in the present research did not affect the pattern of findings. Fourth, although ‘parental mental health’ and ‘child behaviour problems’ in this study are conceptualized as an environmental risk factor, they could also operate as genetic liabilities (Althoff, Rettew, Faraone, Boomsma, & Hudziak, 2006). Separating out the mechanisms by which these variables might operate could not be addressed within a cross-sectional design study, and should be a focus for future studies. Finally, the sample is necessarily a somewhat select one consisting of parents/caretakers who volunteered and who were almost exclusively white British.

While the findings from the present study lend support to a diathesis stress model of sibling adjustment and relationships, it should be noted that only two of the 20 interactions tested led to statistically significant findings. Thus, these results do not currently constitute robust evidence in support of the model. Given the number of analyses conducted, these effects may also represent Type I errors. The aforementioned methodological limitations also raise questions

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555554

regarding the significant interactions that were identified. Future replication of this research may help to determine if the findings were spurious, and to provide additional information about the utility of the diathesis-stress model in this context.

4.1. Implications for research and practice

The findings from the present research suggest that siblings of children with autism who have features characteristic of autism themselves may find it especially difficult to adjust to some environmental stressors in the home. When their siblings have significant behaviour problems and their parents have mental health difficulties siblings with a significant BAP may struggle to adjust, and this is reflected in their own behavioural adjustment and their relationship with their sibling. A negative family environment, in the form of significant levels of parental criticism may also place siblings at risk for poorer sibling relationships. The implications of these results are twofold: for the identification of siblings at risk, and the design of sibling support interventions.

‘Child behaviour problems’ is an environmental risk factor that has been reliably shown to affect sibling adjustment (Hastings, 2006). Parental mental health problems may represent both a direct as well as an indirect environmental risk factor (for example via adversely affecting marital discord, or the emotional/behavioural adjustment of the child with an ASD), in addition to representing a putative genetic risk. Previous studies with siblings of children with an ASD have typically overlooked the moderating effect of the BAP and as such have tended to report directly on the relationship between risk factors and adjustment outcomes, yielding a high number of associations, as was the case with our regression models if one considers the main effects rather than the interaction effects. Such findings tell us little about the pathways of influence of sibling adjustment. Future research should continue to explore the effects of moderating and mediating factors.

In terms of identifying siblings of children with an ASD who may be at risk of problems themselves, practitioners might consider systematic measurement of sibling BAP using a reliable tool. Similarly, when a critical family environment, parent mental health problems, or an ASD child with problem behaviours are present in the family, practitioners should consider the potential risks to siblings. Siblings’ genetic predisposition may manifest as emotional and behavioural difficulties under certain environmental stressors, such as when the child with an ASD has behaviour problems, and/or there is poor parental mental health. A thorough clinical assessment aiming to identify siblings’ needs should therefore examine multiple areas, including individual-constellation, environmental and systemic, as well as developmental and genetic influences.

Developing resilience in the form of social skills and stress coping skills may be an important intervention target for siblings. It is also possible that psychoeducational support would be of assistance. Specifically, increased understanding of autism and associated difficulties (e.g., behaviour problems) and the potential impact on their parents’ health, may help siblings to better adjust when members of their family are suffering significantly.

Our results further stress the importance of family-centred approaches to clinical intervention because of the inter- relationships between the well-being of different family members (also see Hastings et al., 2005). Clinical assessment and intervention should focus on the family system rather than individual members, taking into account the interconnectedness of family subsystems. The present results suggest that siblings are likely to benefit from interventions not necessarily aimed directly and solely at them. Thus, improving behaviour problems in the child with an ASD, treating parental mental health problems such as anxiety or depression, and promoting positive parent–child interactions, might all have benefits for sibling well-being and positive sibling relationships.

Acknowledgements

This research was supported in part by funding from the National Autistic Society Cymru and from the European Social Fund, Objective 1 area funding.

References

Achenbach, T. M. (1991). Manual for the child behaviour checklist. Department of Psychiatry: University of Vermont. Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Althoff, R. R., Rettew, D. C., Faraone, S. V., Boomsma, D. I., & Hudziak, J. J. (2006). Latent class analysis shows strong heritability of the Child Behaviour Checklist –

Juvenile bipolar phenotype. Biological Psychiatry, 60, 903–911. Asarnow, J. R., Tompson, M., Woo, S., & Cantwell, D. P. (2001). Is expressed emotion a specific risk factor for depression or a non-specific correlate of

psychopathology? Journal of Abnormal Child Psychology, 29, 573–583. Auyeung, B., Baron-Cohen, S., Wheelwright, S., & Allison, C. (2007). The Autism Spectrum Quotient: Children’s version (AQ-child). Journal of Autism and

Developmental Disorders, 38, 1230–1240. Baron-Cohen, S., Hoekstra, R. A., Knickmeyer, R., & Wheelwright, S. (2006). The autism-spectrum quotient (AQ) – Adolescent version. Journal of Autism and

Developmental Disorders, 36, 343–350. Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The Autism-Spectrum Quotient (AQ): Evidence from Asperger Syndrome/High-

Functioning Autism, Males and Females, Scientists and Mathematicians. Journal of Autism and Developmental Disorders, 31, 5–17. Bauminger, N., & Yirmiya, N. (2001). The functioning and well-being of siblings of children with autism: Behavioral-genetic and familial contributions. In J. A.

Burack, T. Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The development of autism: Perspectives from theory and research (pp. 61–80). Mahwah, NJ: Lawrence Erlbaum.

Benson, P. R., & Karlof, K. L. (2008). Child, parent and family predictors of later adjustment in siblings of children with autism. Research in Autism Spectrum Disorders, 2, 583–600.

Bjelland, I., Dahl, A. A., Haug, T. T., & Necklemann, D. (2002). The validity of the hospital anxiety and depression scale: An updated literature review. Journal of Psychosomatic Research, 52, 69–77.

M.A. Petalas et al. / Research in Autism Spectrum Disorders 6 (2012) 546–555 555

Bolton, P. F., Pickles, A., Murphy, A., & Rutter, M. (1998). Autism, affective and psychiatric disorders: Patterns of familial aggregation. Psychological Medicine, 28, 385–395.

Buhrmester, D., & Furman, W. (1990). Perceptions of sibling relationships during middle childhood and adolescence. Child Development, 61, 1387–1389. Conger, R. D., Wallace, L. E., Sun, Y., Simons, R. L., McLoyd, V. C., & Brody, G. H. (2002). Economic pressure in African American families: A replication and extension

of the family stress model. Developmental Psychology, 38, 179–193. Derkman, M. S., Scholte, R. H. J., Van der Veld, W. M., & Engels, R. C. M. E. (2010). Factorial and construct validity of the sibling relationship questionnaire. European

Journal of Psychological Assessment, 26, 277–283. Dunn, J., Slomkowski, C., Beardsall, L., & Rende, R. (1994). Adjustment in middle childhood and early adolescence: Links with earlier and contemporary sibling

relationships. Journal of Child Psychology and Psychiatry, 35, 491–504. Elander, J., & Rutter, M. (1996). Use and development of the Rutter parents’ and teachers’ scales. International Journal of Methods in Psychiatric Research, 6, 63–78. Ferrari, M. (1984). Chronic illness: Psychosocial effects on siblings – I. Chronically ill boys. Journal of Child Psychology and Psychiatry, 25, 459–476. Fisman, S., Wolf, L., Ellison, D., Gillis, B., Freeman, T., & Szatmari, P. (1996). Risk and protective factors affecting the adjustment of siblings of children with chronic

disabilities. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 1532–1541. Goodman, R. (1997). The strengths and difficulties questionnaire. Journal of Child Psychology and Psychiatry, 40, 791–799. Goodman, R., & Scott, S. (1999). Comparing the strengths and difficulties questionnaire and the child behaviour checklist: Is small beautiful? Journal of Abnormal

Child Psychology, 27, 17–24. Greenberg, J. S., Seltzer, M. M., Hong, J., & Orsmond, G. L. (2006). Bidirectional effects of expressed emotion and behavior problems and symptoms in adolescents

and adults with autism. American Journal on Mental Retardation, 111, 229–249. Hastings, R. P. (2006). Longitudinal relationships between sibling behavioral adjustment and behavior problems of children with developmental disabilities.

Journal of Autism and Developmental Disorders, 37, 1485–1492. Hastings, R. P. (2007). Longitudinal relationships between sibling behavioral adjustment and behavior problems of children with developmental disabilities.

Journal of Autism and Developmental Disorders, 37, 1485–1492. Hastings, R. P., & Lloyd, T. (2007). Expressed emotion in families of children and adults with intellectual disabilities. Mental Retardation and Developmental

Disabilities Research Reviews, 13, 339–345. Hastings, R. P., Kovshoff, H., Ward, N. J., Espinosa, F. D., Brown, T., & Remington, B. (2005). Systems analysis of stress and positive perceptions in mothers and

fathers of pre-school children with autism. Journal of Autism and Developmental Disorders, 35, 635–644. Kaminsky, L., & Dewey, D. (2001). Sibling relationships of children with autism. Journal of Autism and Developmental Disorders, 33, 1679–1690. Magana, A. B., Goldstein, M. J., Karno, M., Miklowitz, D. J., Jenkins, J., & Fallon, I. R. H. (1986). A brief method for assessing expressed emotion in relatives of

psychiatric patients. Psychiatry Research, 17, 203–212. McCarthy, C. A., Lau, A. S., Valeri, S. M., & Weisz, J. R. (2004). Parent–child interactions in relation to critical and emotionally over involved expressed emotion (EE):

Is EE a proxy for behavior? Journal of Abnormal Child Psychology, 32, 83–93. Orsmond, G. I., Kuo, H., & Seltzer, M. M. (2009). Siblings of individuals with an autism spectrum disorder: Sibling relationships and well-being in adolescence and

adulthood. Autism, 13, 59–80. Orsmond, G. I., & Seltzer, M. M. (2009). Adolescent siblings of individuals with autism spectrum disorder: Testing a diathesis-stress model of sibling well-being.

Journal of Autism and Developmental Disorders, 39, 1053–1065. Orsmond, G. I., Seltzer, M. M., Greenberg, J. S., & Krauss, M. W. (2006). Mother–child relationship quality among adolescents and adults with autism. American

Journal of Mental Retardation, 111, 121–137. Petalas, M. A., Hastings, R. P., Nash, S., Lloyd, T., & Dowey, A. (2009). Emotional and behavioural adjustment in siblings of children with intellectual disability with

and without autism. Autism, 13, 471–483. Pike, A., Coldwell, J., & Dunn, J. F. (2005). Sibling relationships in early/middle childhood: Links with individual adjustment. Journal of Family Psychology, 19, 523–

532. Slomkowski, C., Rende, R., Conger, K. J., Simons, R. L., & Conger, R. D. (2001). Sisters, brothers, and delinquency: Evaluating social influence during early and middle

adolescence. Child Development, 72, 271–283. Wolf, L., Fisman, S., Ellison, D., & Freeman, M. D. (1998). Effect of sibling perception of differential parental treatment in sibling dyads with one disabled child.

Journal of American, Academic Child and Adolescent Psychiatry, 37, 1317–1325. Yan, L. J., Hammen, C., Cohen, A. N., Daley, S. E., & Henry, R. M. (2004). Expressed emotion versus relationship quality variables in the prediction of recurrence in

bipolar patients. Journal of Affective Disorders, 83, 199–206. Zigmond, A. S., & Snaith, R. P. (1983). The Hospital Anxiety and Depression Scale. Acta Psychiatrica Scandinavica, 67, 361–370.