Quality of interaction between at-risk infants and caregiver at months is associated with 3-year autism outcome

Journal of Child Psychology and Psychiatry 54:7 (2013), pp 763 771 doi:10.1111/jcpp.12032 Quality of interaction between at-risk infants and caregiver at 12 15 months is associated with 3-year autism outcome Ming Wai Wan, 1 Jonathan Green, 1 Mayada Elsabbagh, 2 Mark Johnson, 2 Tony Charman, 3 Faye Plummer, 1 and the BASIS Team* 1 Institute of Brain, Behaviour and Mental Health, University of Manchester, Manchester; 2 Centre for Brain and Cognitive Development, Birkbeck, University of London, London; 3 Centre for Research in Autism and Education, Institute of Education, University of London, London, UK Background: Recent models of the early emergence of autism spectrum disorder (ASD) propose that infant intrinsic risk susceptibilities in behaviour may be amplified by interaction within the early social environment into an increasingly atypical developmental trajectory. This study examines whether 6- and 12-month parent infant interactions in at-risk siblings differ from those with low-risk and whether in at-risk siblings such interactions predict later 3-year classification of ASD or no ASD. Method: Within the British Autism Study of Siblings (BASIS), 6-min videotaped episodes of parent infant free play in infants at 6 10 months (45 at-risk siblings and 47 low-risk siblings) and 12 15 months (43 at-risk siblings and 48 low-risk siblings) in a laboratory setting were rated on the Manchester Assessment of Caregiver- Interaction (MACI), blind to participant information. Standard tests were administered for concurrent behavioural signs of ASD features and developmental level. Systematic consensus diagnostic classification of ASD was made at 3 years for the at-risk siblings. Results: Parent nondirectiveness and sensitive responsiveness differed in relation to ASD/risk status (at-risk ASD, at-risk no-asd and low-risk) at both 6 and 12 months. At 6 months, infant liveliness was lower in the at-risk groups; at 12 months, infant attentiveness to parent and positive affect were lower in the at-risk group later diagnosed with ASD. Dyadic mutuality and intensity of engagement showed a group effect at 12 months. Dyadic mutuality, infant positive affect and infant attentiveness to parent at 12 months (but not 6 months) predicted 3-year ASD outcome, whereas infant ASD-related behavioural atypicality did not. Conclusions: This is the first prospective evidence that early dyadic interaction between at-risk infants and their parents is associated with later diagnostic outcome in ASD. Possible explanations for these findings and their theoretical implications are considered.keywords: ASD siblings, mother child relations, high-risk infants, parent sensitivity. Introduction Overt behavioural signs of autism spectrum disorder (ASD) are rarely observable in the first year, but increasing evidence suggests that studying early caregiver infant interaction in the context of emergent ASD may be important for a more complete understanding of the development of the disorder (Dawson, 2008; Wallace & Rogers, 2010). A transactional model posits that early social experience shapes long-term social functioning (Sameroff, 2009; Sameroff & Fiese, 1990). In normative development, the quality of interaction between primary caregiver and infant (i.e. the infant s main early social environment) influences the infant s own social developmental trajectory (e.g. Feldman & Greenbaum, 1997; NICHD Early Child Care Research Network, 2001). In the context of ASD, this does not * The BASIS team in alphabetical order: Simon Baron-Cohen, Rachael Bedford, Patrick Bolton, Susie Chandler, Janice Fernandes, Holly Garwood, Teodora Gliga, Kristelle Hudry, Leslie Tucker, Greg Pasco and Agnes Volein. Conflict of interest statement: The authors declare no conflicts of interest. suggest that parents in any way cause the disorder. Rather, according to the transactional model, any interactive perturbations whether originating in the infant or caregiver s behaviour may become increasingly entrenched or amplified with time, impacting on later social communicative functioning. There are three primary explanations as to why early parent infant interactive perturbations might arise in the context of studying emergent ASD, none of which are mutually exclusive. First, interaction (including parental response) may primarily reflect the infant s emerging behavioural and social atypicalities in those infants who go onto develop ASD. Second, it is possible that parents of infants at greater familial risk of ASD are more likely to display partial ASD characteristics or broader autism phenotypic (BAP) impairments (e.g. Losh, Childress, Lam, & Piven, 2008), which may impact on interaction. The third possibility arises from the at-risk prospective study design used to investigate prodromal ASD, which involves infants who have an older sibling with ASD ( at-risk siblings ), a proportion of whom will develop the disorder because of Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA

764 Ming Wai Wan et al. J Child Psychol Psychiatry 2013; 54(7): 763 71 familial risk. In this context, interaction may be disrupted in at-risk siblings as a consequence of parents learned styles of interacting with the older child with ASD, which may differently affect interaction with their younger sibling without ASD. A few studies have examined the parent infant play interactions of at-risk siblings compared with control infants (without a sibling with ASD low-risk siblings ), and less commonly, of at-risk siblings who are later diagnosed. Studies have tended to show specific tendencies in parent interactive behaviour: The earliest study was a microanalysis of 4-monthold at-risk siblings (n = 21), which found that a significant proportion had less affectively synchronous interactions when infants led play compared with low-risk comparisons, suggesting that these mothers find it difficult to match infant-initiated affect (Yirmiya et al., 2006). Parental play interactions in our larger sample of at-risk infants (n = 45; 6 10 months) were rated to be more directive and less sensitively responsive than found in low-risk controls (Wan et al., 2012). This finding is consistent with a retrospective study of parents home videos of infants in the same age range, which found that specific directive behaviours (longer stimulatory behaviour and more use of touch to elicit attention) differentiated parents whose infants were later diagnosed with ASD (n = 15) from parents of typically developing infants and infants with intellectual disabilities (Saint-Georges et al., 2011). There has so far been less evidence for disrupted interactive behaviour in at-risk infant siblings, despite suggestive findings from naturalistic retrospective studies (Saint-Georges et al., 2010). At-risk siblings were less lively than low-risk controls in Wan et al. s (2012) interaction study, although liveliness was not associated with other qualitative aspects of interaction. One retrospective home movie study that focused on parent infant interaction found lower infant-initiated orientation to the parent at 6 10 months than in groups who were typically developing or had intellectual disabilities (Saint-Georges et al., 2011). By contrast, at-risk sibling studies have found no infant social orientation differences during interaction measured either globally (Wan et al., 2012) or microanalytically in at-risk siblings who develop ASD (Rozga et al., 2011). Taking a transactional model, if parent infant interaction forms a pattern that contributes to social development, then we might expect first that such interaction is usually relatively stable, and second that any perturbations in interaction would amplify over time and impact on infant social functioning a key deficit in ASD. However, no interaction studies so far have found any such longer-term impact. The lower infant-led synchrony identified by Yirmiya et al. (2006) was unrelated to 14-month language and communication. However, in a study of older atrisk siblings, Baker, Messinger, Lyons, and Grantz (2010) found that positive parental structuring (but not sensitivity generally) during 18-month interaction was positively related to expressive language change in those who went on to develop ASD. This study followed up parent infant interaction in an at-risk sibling cohort (Wan et al., 2012) from 6 10 months( 6 months ) to12 15 months( 12 months ) in relation to 3-year ASD outcome. Ascribing to a transactional model that early interaction quality is crucial for longer-term social functioning and outcome, it was predicted that parent infant interaction (a) would show moderate stability and internal consistency between 6 and 12 months in low-risk siblings, (b) would differ significantly in at-risk siblings who are later diagnosed with ASD from those who are not and from low-risk siblings at 6 months and (more so, as differences become amplified) at 12 months, and (c) would be predictive in at-risk siblings of 3-year ASD classification, independent of the level of early behavioural atypicality (as measured through structured assessment). That is, interaction effects are not simply explained by behavioural atypicality. To further disentangle how parent infant interaction might relate to ASD markers, these variables were also explored at 12 months in at-risk siblings who were subthreshold for an ASD diagnosis at 3 years. Method Sample Recruitment, ethical approval (UK National Health Service National Research Ethics Service London ref: 09/H0718/14), informed consent and background data on participating families were made available for this study through the British Autism Study of Siblings (BASIS), a UK collaborative network facilitating research with infants at risk for ASD (http:// www.basisnetwork.org). At-risk sibling status was defined by having an older sibling with a clinical diagnosis of ASD (or in 4 cases, a half-sibling), confirmed by two expert clinicians (PB and TC) based mainly on information using the Development and Wellbeing Assessment (DAWBA; Goodman, Ford, Richards, Gatward, & Meltzer, 2000) and the parent-report Social Communication Questionnaire (SCQ; Rutter, Bailey, & Lord, 2003). s in the low-risk group were recruited separately from a volunteer database. Inclusion criteria included full-term birth, normal birthweight and lack of ASD within first-degree family members (confirmed through parent interview). Low-risk infants had at least one older sibling, scoring below the SCQ threshold (<15) for ASD. Of 104 dyads recruited (54 at-risk siblings and 50 low-risk siblings), 12 were excluded from the 6-month cohort and 12 from the 12-month cohort, as interaction data were not collected or because of technical issues. Of the remainder, one further case was excluded from the 12-month sample because of lack of assessment at 36 months. Complete 6-month data were available for 45 at-risk infants (20 male, 44.4%) and 47 low-risk infants (18 male, 38.3%), and 12-month data for 43

doi:10.1111/jcpp.12032 At-risk infant interaction and autism 765 at-risk infants (17 male, 39.5%) and 48 low-risk infants (17 male, 35.4%). Interaction data at both time points were taken for 81 dyads (39 at-risk siblings and 42 lowrisk siblings). All parents who took part in the interactions were mothers (See online Appendix for further details of the sample). Measures Parent infant interaction: A global rating scheme, the Manchester Assessment of Caregiver- Interaction (MACI), was used to evaluate seven items (on a 1 7 scale): caregiver sensitive responsiveness, caregiver nondirectiveness, infant attentiveness to caregiver, infant positive affect, infant liveliness, dyadic mutuality and dyadic intensity of engagement (Details in online Appendix). The measure is an adaptation of existing global scales of caregiver infant interaction for use in this age range and focus of study. Excellent psychometric properties and inter-rater reliability were reported in our previous study at 6 10 months (Wan et al., 2012), where ratings were independent of infant gender, infant nonverbal development, maternal age and socioeconomic status. The scales were further piloted for use in 12 to 15-month-old infants and minor amendments made to ensure a distribution of ratings. Independently blind-rated clips (26%) showed reasonable to high agreement (single measures intraclass correlations using a two-way mixed effects model; absolute agreement) ranging from r = 0.68 to r = 0.83 (p < 0.001; see Appendix for further details). Atypical infant behaviour: behaviours characteristic of emergent autism were assessed using the Autism Observation Scale for s (AOSI; Bryson, Zwaigenbaum, McDermott, Rombough, & Brian, 2008), a validated clinical measure of infant risk markers, focusing on precursors of ASD phenotype impairments, including response to name, eye contact, social reciprocity and imitation. behaviour is elicited, while on the parent s lap, within a structured interaction with an assessor including a series of social presses. A 7+ behavioural marker threshold in AOSI at 1 year has predicted ASD diagnosis at 24 months with a sensitivity of 84% and specificity of 98% (Bryson et al., 2008). The AOSI was coded by the administrator and (in most cases) another trained researcher, both trained and supported in reliability by one of the originators of the instrument. Inter-rater reliability for 1-year AOSI total score was excellent (single measures, absolute agreement; N = 96; r = 0.92; p < 0.001). Discrepancies were resolved by consensus and re-reviewing performance on videotape. developmental level: The Mullen Scales of Early Learning (MSEL; Mullen, 1995) is a standardised test of early receptive and expressive language, visual reception and gross and fine motor skills for use in 0 to 68-month children, of which the early learning composite (ELC) standard scores was used. Autism outcome at 3 years: Following the 36- month assessment in the at-risk group, an independent team used triangulation of all available information from all visits, combined with expert clinical judgment (TC, KH, SC and GP), to determine a consensus ICD-10 (World Health Organisation, 1993) Autism spectrum disorder classification, including childhood ASD, atypical ASD and other pervasive developmental disorder. Information used to determine possible diagnosis included 24- and 36-month results from the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 1989). This play-based assessment rates from videotape a child s social behaviour in the context of a number of social presses administered by a trained assessor and designed to elicit reciprocal social interaction, language and communication and repetitive and stereotyped behaviours. At 36 months, one at-risk sibling was administered ADOS module 1 and 89 module 2 (two missing data including one at-risk sibling). ADOS assessment was augmented with the parent-report ASD Diagnostic Interview-Revised (Lord et al., 1994). Given the children s young age, and in line with the proposed changes to DSM-5, no attempt was made to assign specific subcategories of PDD/ASD diagnosis. Within the 6-month at-risk sibling cohort, of those 45 assessed at 36 months, 14 [31.1%; 10 (17.4%) boys] met criteria for ASD diagnosis ( at-risk ASD ) and 31 did not [ at-risk no-asd ; 10 boys (32.3%)]. In the 12-month at-risk sibling cohort, 12 [27.9%; eight boys (66.7%)] met criteria for an ASD diagnosis and 31 [72.1%; nine boys (29.0%)] did not. Within the 12-month cohort, 11 at-risk group toddlers (25.6%; three boys and eight girls) were classified as not having ASD but considered to have partial atypicality or other developmental concerns by virtue of their profile of assessment scores. Nine of this group scored >ADOS cut-off, one >ADOS cut-off and <1.5 SD MSEL ELC cut-off, and one <1.5 SD MSEL ELC cut-off. Procedure At the 12-month laboratory assessment, an episode of play interaction between parent and infant was video recorded, following a period of familiarisation. The parent was instructed to engage in play as she would do at home, using the set of toys provided if she wished. Clips were later rated (the first 6 min from the point the researchers left the room) independently, blind to participant information. During the same assessment visit, the MSEL and AOSI were administered by a trained researcher, independent of interaction ratings. Data analysis Separately for at-risk and low-risk sibling groups, two sets of correlations were calculated: (a) between MACI rating scales at 12 months to examine internal consistency (see Wan et al., 2012 for 6-month analyses) and (b) between time points (6 and 12 months) on each MACI scale to examine stability, based on the availability of interaction data at both time points (N = 81; 10 missing). Owing to this reduced sample, subsequent analyses were completed separately for each time point. Group differences in parent infant interaction by 3-year ASD/risk status (three groups: at-risk ASD, atrisk no-asd and low-risk) were examined using oneway ANOVA followed by ANCOVA adjusting for infant

766 Ming Wai Wan et al. J Child Psychol Psychiatry 2013; 54(7): 763 71 age. Main analyses did not adjust for MSEL, based on Dennis et al. s (2009) logic that in young children who do not have severe learning disability (as in this sample), measurement of developmental level is almost certain to include elements of emergent ASD that make this measurement impossible to disentangle from the developmental disorder itself, thus including MSEL in such analyses runs the risk of over-adjustment. The 12-month interaction data from a proportion of children who received no ASD diagnosis but showed other concerns (see Measures) was analysed descriptively, as the sample size did not allow for testing predictions of outcome in a four-way classification. Finally, separately at 6 and 12 months, those interaction domains in which an effect by sibling/asd status was found after controlling for age were examined using binomial logistic regression in relation to later ASD classification. Owing to their interrelated nature, statistically and conceptually, each interaction domain was tested in a separate regression model. As a first step in this analysis, 12-month AOSI score was entered in a univariate model to test the variance this accounted for in ASD classification. Next, relevant interaction variables were entered in separate models with age as covariate. Results Sample characteristics At both time points, no differences were found by ASD/risk status in infant gender, infant age or parental age, but relative to the low-risk group, a significantly higher proportion of parents of at-risk siblings (combined for chi-square because of sample size) had infants with >1 sibling and came from professional/managerial occupations (Table 1). However, it was the at-risk ASD group rather than those who did not receive a diagnosis who had more similar family backgrounds to the low-risk group. Significant group effect emerged in the AOSI at 6 months (at 12 months, the number of markers showed a borderline effect: p = 0.06), which Scheffe post hoc tests found was due mainly to the difference between the at-risk ASD group from the low-risk group, and in the MSEL ELC, which Scheffe post hoc tests attributed to the high score in the low-risk group at 6 months (compared with both other groups), and the low score in the at-risk ASD group at 12 months. Parent infant interaction: 12-month characteristics and 6 to 12-month stability At 12 months, most areas of interaction were significantly intercorrelated in both risk groups (Table 2). liveliness was unrelated to other domains in both groups, except for a negative correlation with parent nondirectiveness in the low-risk group. positive affect was highly correlated with other domains in the low-risk group but, within at-risk siblings, was not associated with parent domains, particularly nondirectiveness. Twelvemonth interaction showed no concurrent association with infant age, AOSI or parental occupation. While ELC did not correlate with interaction in the at-risk group (combined, because of sample size), it did so in the low-risk group with several areas of interaction (infant attentiveness: r = 0.30; mutuality: r = 0.29; intensity of engagement: r = 0.35; all p 0.05). Based on the cohort for which interaction data were available across time points (N = 39 at-risk siblings; 42 low-risk siblings), and controlling for infant age, moderate stability was observed in the low-risk group in the parent domains, infant Table 1 Descriptive characteristics by sibling group at 6 and 12 months 6 months 12 months Sibling group mean (SD) Sibling group mean (SD) At-risk ASD (N = 14) At-risk no-asd (N = 31) Low-risk (N = 47) F At-risk ASD (N = 12) At-risk no-asd (N = 31) Low-risk (N = 48) F age (months) 7.28 (1.18) 7.10 (1.22) 7.36 (1.19) 0.47 14.00 (1.21) 13.35 (1.38) 13.92 (1.56) 1.62 Parent age (years) 35.14 (5.04) 36.35 (4.27) 36.40 (4.66) 0.44 35.92 (4.81) 36.81 (4.22) 36.81 (4.59) 0.21 MSEL ELC 87.92 (11.77) 93.77 (10.31) 103.81 (10.76) 14.96** 88.25 (16.77) 102.45 (16.82) 106.11 (15.73) a 5.78** AOSI total score 10.36 (5.65) 8.00 (4.52) 6.55 (3.83) 4.26* 5.75 (3.84) 4.13 (4.63) 3.17 (3.25) 2.30 AOSI markers 6.00 (3.04) 5.13 (2.39) 3.83 (2.20) 5.54** 3.58 (2.19) 2.87 (2.87) 2.04 (1.68) 2.89 N within sibling group (%) N within sibling group (%) At-risk ASD (N = 14) At-risk no-asd (N = 31) Low-risk (N = 45) v 2b At-risk ASD (N = 12) At-risk no-asd (N = 31) Low-risk (N = 48) v 2b has >1 sibling 7 (50.0%) 20 (64.5%) 13 (28.9%) 9.32** 5 (41.7%) 18 (58.1%) 14 (29.2%) 5.10** Parent professional/ 11 (84.6%) 18 (62.1%) 41 (87.2%) 7.09* 9 (75.0%) 20 (66.7%) 42 (87.5%) 4.58* managerial a *p < 0.05; **p < 0.01. a Missing data: 6 months: one at-risk ASD, two at-risk no-asd; 12 months: one at-risk no-asd. b Chi-squared analysis is based on the combined at-risk group, because of low expected cell count.

doi:10.1111/jcpp.12032 At-risk infant interaction and autism 767 Table 2 Correlations between 12-month global aspects of interaction in at-risk and low-risk siblings a Parent sensitive responsive Parent nondirectiveness attentiveness positive affect liveliness Mutuality Intensity of engagement At-risk siblings (N = 44) Sen responsiveness.63**.60**.28 (p = 0.07) ).09.78**.71** Nondirectiveness.60**.45*.14 ).07.54**.44** Attentiveness.80**.37**.34* ).25.87**.70** Positive affect.42**.29*.39**.18.29**.43** Liveliness ).28 ).43** ).26.01 ).23 ).15 Mutuality.93**.49**.87**.44**.26.79** Engagement.77**.34*.84**.46** ).21.83** Low-risk siblings (N = 48) *p < 0.05 ** p < 0.01. a Adjusting for infant age and MSEL yielded very similar results, although in the low-risk group, the correlation between infant liveliness and attentiveness became significant (r = ).28; p = 0.05). Table 3 Consistency of parent infant interaction between 6 and 12 months: partial correlations controlling for infant age Sibling group Parent sensitive responsive Parent nondirectiveness attentiveness positive affect liveliness Mutuality Intensity of engagement At-risk (N = 39).24.23.24.18.29.27.43** Low-risk (N = 45) 48**.37*.37* ).06 ).11.46**.27 p < 0.1; *p < 0.05; **p < 0.01. attentiveness and mutuality, and in the at-risk group in intensity of engagement only (Table 3). Six- and 12-month parent infant interaction by 3-year ASD/risk status Group differences by ASD/risk status were found in parent interactive behaviours, with borderline significant effects in sensitive responsiveness at both time points and stronger effects in nondirectiveness (Table 4). Whereas at 6 months, these ratings were lower in the at-risk sibling groups than in the lowrisk group irrespective of later ASD status, at 12 months, the at-risk ASD group specifically showed lower ratings compared with the low-risk group as confirmed by Scheffe post hoc tests. Table 4 Global ratings of parent infant interaction by ASD/risk status at 6 months and 12 months Sibling group mean (SD) At-risk ASD At-risk No-ASD Low-risk 6 months N =14 N =31 N =47 Unadjusted F (p value) Adjusted F for infant age (p value) Parent scales Sensitive responsiveness 3.14 (0.95) 3.23 (1.23) 3.79 (1.33) 2.56 (0.08) 2.23 (0.11) Nondirectiveness 3.14 (1.35) 3.03 (1.28) 3.91 (1.33) 4.80 (0.01)** 4.35 (0.02)* scales Attentiveness to parent 3.71 (1.64) 3.90 (1.19) 3.89 (1.39) 0.11 (0.90) 0.14 (0.87) Positive affect 4.36 (0.84) 3.77 (1.26) 3.79 (1.02) 1.67 (0.19) 1.77 (0.18) Liveliness 3.69 (0.86) 3.42 (1.23) 4.29 (1.06) 6.03 (0.004)** 5.66 (0.005)** Dyad scales Mutuality 3.00 (1.52) 3.03 (1.28) 3.17 (1.42) 0.13 (0.88) 0.09 (0.92) Engagement intensity 3.86 (1.46) 4.93 (1.22) 3.91 (1.33) 0.11 (0.90) 0.15 (0.86) 12 months N =12 N =31 N =48 Parent scales Sensitive responsiveness 2.92 (1.08) 3.58 (1.41) 3.98 (1.50) 2.86 (0.06) 2.83 (0.07) Nondirectiveness 3.17 (1.34) 3.55 (1.52) 4.31 (1.43) 4.37 (0.02)* 4.03 (0.02)* scales Attentiveness to parent 3.17 (1.19) 4.37 (1.22) 4.67 (1.28) 6.79 (0.002)** 6.95 (0.002)** Positive affect 3.00 (0.74) 4.06 (1.00) 4.04 (0.85) 7.26 (0.001)** 7.31 (0.001)** Liveliness 5.33 (1.07) 4.77 (1.04) 5.00 (1.07) 0.91 (0.41) 0.75 (0.48) Dyad scales Mutuality 2.25 (0.87) 3.61 (1.50) 3.92 (1.54) 6.27 (0.003)** 6.50 (0.002)** Engagement intensity 3.42 (0.90) 4.40 (0.93) 4.21 (0.74) 4.45 (0.01)** 4.35 (0.02)* p < 0.1; *p < 0.05; **p < 0.01.

768 Ming Wai Wan et al. J Child Psychol Psychiatry 2013; 54(7): 763 71 In infant interactive behaviours, a significant group effect was found only in liveliness at 6 months, and at 12 months in attentiveness to parent and positive affect. Scheffe post hoc tests suggest that the latter effects were due mainly to the low scores of the at-risk ASD group compared with both other groups. Twelve-month dyadic group differences also emerged in both mutuality and intensity of engagement. Scheffe post hoc tests again located the effect specifically to low ratings in the atrisk ASD group. The 12-month data of the other concerns subgroup (see Method) were further explored descriptively and showed a distinct pattern, consisting of elevated mean AOSI scores similar to the ASD group combined with positive interaction ratings more similar to the non-asd outcome groups. Ratings in all six interaction domains showing main group effects were similar to the low-risk group and at-risk no-asd group (see Appendix for details). ASD diagnostic outcome: The predictive value of AOSI and interaction Table 5 Logistic regression models of the independent contributions of 1-year AOSI, infant age and domains of parent infant interaction to 3-year diagnostic classification Model Predictor Wald (p value) Nagelkerke R 2 1 AOSI total score 1.13 (0.29) 0.04 2 AOSI total score 0.70 (0.40) 0.10 age (months) 0.05 (0.83) Parent sensitive 1.67 (0.20) responsiveness 3 AOSI total score 1.53 (0.22) 0.08 age (months) 0.09 (0.77) Parent nondirectiveness 1.05 (0.30) 4 AOSI total score 0.55 (0.46) 0.24 age (months) 0.01 (0.94) attentiveness 4.97 (0.03)* to parent 5 AOSI total score 0.48 (0.49) 0.39 age (months) 1.74 (0.18) positive affect 6.72 (0.01)** 6 AOSI total score 0.59 (0.44) 0.30 age (months) 0.19 (0.66) Dyadic mutuality 5.70 (0.02)* 7 AOSI total score 0.22 (0.64) 0.22 age (months) 1.00 (0.32) Dyadic intensity of engagement 3.09 (0.08) p < 0.1; *p < 0.05; **p < 0.01. At 6 months, neither AOSI nor interaction ratings where main effects had been found predicted significant variance in 36-month outcome (See Appendix). However, at 12 months, three areas of interaction predicted ASD outcome, independent of AOSI and infant age: infant positive affect, dyadic mutuality and infant attentiveness to parent (Table 5, Models 5, 6 and 4, respectively). A borderline significant effect was found in intensity of engagement. Lower ratings in all domains were associated with ASD outcome. The model incorporating infant positive affect accounted for approximately 39% of the variance in ASD outcome. Neither AOSI alone nor specific parent interaction domains accounted for significant variance (Models 1 3). As only infant and dyadic interaction domains (at 12 months) and not parent domains (as at 12 months) were associated with later ASD, similar regression analyses were then conducted with the 6-month data to test for possible earlier effects, but no significant results were found. A re-analysis with MSEL as a further covariate made no essential difference to the results. The same significant interaction predictors were found albeit at a slightly reduced level in two areas (infant attentiveness: wald = 4.19; p = 0.04; mutuality: wald = 4.92; p = 0.03) and increased in infant positive affect (wald = 6.77; p = 0.009; with an independent effect of MSEL ELC: wald = 3.80; p = 0.05; overall Nagelkerke R 2 = 0.50). Discussion This study demonstrates for the first time, through cross-sectional analyses of a cohort of infants at familial risk, parent infant interaction at 6 10 months and 12 15 months, and prediction from 12 to 15-month interaction to ASD outcome at 3 years. Moderate stability across the (reduced) cohort is seen from 6 to 12 months mainly in the low-risk group in parent, infant and dyadic domains of interaction, which showed similar though increasing internal relationships between time points (c. f. Wan et al., 2012). Yet, these cross-sectional findings may suggest an increasingly different trajectory in the interactions of at-risk siblings from controls, and in those who are later diagnosed with ASD from those who are not. At 6 months, group differences in at-risk siblings compared with low-risk controls are largely seen in parent interactive behaviour, as suggested in previous studies (George-Saints et al., 2011; Yirmiya et al., 2006), but are unrelated to subsequent ASD. Thus, our hypothesis that 6- month interaction would predict 3-year ASD classification was not supported, consistent with the general lack of predictive value of infant behavioural markers at 6 months (c. f. Yirmiya & Charman, 2010). However, at 12 months, the group differences between at-risk siblings and controls are more pervasive, with infant and dyadic aspects of interaction affected. A significant relationship is seen between these infant and dyadic aspects of interaction at 12 months and later ASD outcome, providing the first prospective evidence for interactive differences in infancy in those later diagnosed with ASD. These effects are independent of infant behavioural markers measured independently using the AOSI.

doi:10.1111/jcpp.12032 At-risk infant interaction and autism 769 Furthermore, although a difference was found in 6- month AOSI between sibling groups, the predictive validity of AOSI at 12 months to subsequent diagnosis (e.g. Zwaigenbaum, Bryson, Rogers, Roberts, & Szatmari, 2005) was not replicated here. As for why parent infant interaction differs in at-risk siblings who go onto develop ASD from those who do not, our pattern of results can be interpreted in several ways, and the inter-related nature of social interaction makes it difficult to discern which mechanism(s) are involved. As parent interactive behaviours at 6 and 12 months are affected in at-risk siblings irrespective of later ASD outcome, our results may suggest that such interactive differences are not driven primarily by the presence of early ASD-related atypicalities. However, the directive tendencies found in parents of at-risk siblings could suggest a learned directive style from interacting with their older child with ASD, which might at least contribute to our findings. One study of older children found that their mothers seem to differentiate in their responsiveness but not initiatives towards their child with ASD and unaffected sibling (Meirsschaut, Warreyn, & Roeyers, 2011). It is also possible that interaction effects are due to the partial ASD characteristics of parents of at-risk siblings, because parental BAP has been shown to be more prevalent within multiplex families than simplex families (Losh et al., 2008) although some studies have shown effects for fathers only (Schwichtenberg, Young, Sigman, Hutman, & Ozonoff, 2010; Virkud et al., 2009), while our study involved only mothers. Finally, infant attentiveness to parent and affect differences are seen at 12 months but not 6 months, and these predict ASD outcome at 3 years among at-risk siblings, which might suggest that emerging atypicalities related to later ASD may impact on interaction later in infancy. If early emerging atypicalities manifest themselves in disrupted parent infant interaction, then the simplest explanation linking early interaction with ASD outcome is that interaction in this context (i.e. free play with a familiar caregiver) is simply a more accurate or sensitive reflection of emerging atypicality at this age range (than found using more structured measures by an unfamiliar adult), which in turn predicts later ASD diagnosis. The AOSI itself at 12 months was unrelated to 3-year diagnosis. Moreover, a previous longitudinal study found no relationship between 4-month interaction and 14- month structured measures (Yirmiya et al., 2006). This would suggest that more naturalistic measures of interaction are needed. However, a second interpretation is that whatever the primary cause(s) of disrupted interaction in atrisk siblings, they are likely to reduce social learning opportunities for the infant which, when combined with possible early functional brain differences (e.g. Elsabbagh et al., 2012; Key & Stone, 2012), might affect longer-term social functioning. Taking this transactional approach, these effects in at-risk infants may bias the child towards a more atypical developmental trajectory and ASD classification. In support of this, our findings suggest amplifying interaction difficulties over time leading to greater impact on outcome. attentiveness, infant affect and dyadic mutuality at 1 year, but not at 6 months, show independent associations with later ASD outcome. Furthermore, although speculative at this stage, our other concerns group descriptive findings could suggest that positive interaction might move social development into a more typical developmental trajectory. Whereas at-risk siblings with relatively high AOSI scores at 1 year who had low interaction ratings tended to develop ASD, those high AOSI scorers with more social interactions (similar to the low-risk group) went onto have a non-asd other concerns outcome. Some limitations in design and sampling need to be considered in the interpretation of results. The at-risk sample differed from the low-risk group in family size and socioeconomic status; however, neither variable was related to interaction. Further analyses adjusting for parental occupation (not reported here due to space) found that the group effects held. The recurrence rate reported (32.1%) is higher than found by others (e.g. Ozonoff et al., 2011), probably reflecting the at-risk sample size; these rates are sample specific and are not intended to be representative. Our sample was still significantly larger than previous studies examining early parent infant interaction in relation to ASD and is adequately powered for the analyses undertaken. However, a larger sample with more repeated measures into toddlerhood is needed to delineate precisely the developmental processes leading to the presence or absence of ASD in these at-risk groups. There were various reasons as to why we were unable to obtain codeable parent infant interaction data at both time points for some of the sample (e.g. technical issues), which unfortunately precluded the longitudinal measurement that is necessary to confirm our interpretation regarding the developmental progression of parent infant interaction in the context of ASD risk. No measures of parental BAP or psychopathology were taken, which might reasonably have impacted on interaction quality, although there is no evidence that parental psychopathology might bias towards emergent ASD. A microanalytic coding approach might also allow better delineation of the processes involved (for example, what aspects of infant inattentiveness might impact on parent behaviour and how these relate to diagnosis). However, ultimately, making causal inferences about the effect of early interaction on later ASD development will be best established by intervention designs using random allocation to test the effects

770 Ming Wai Wan et al. J Child Psychol Psychiatry 2013; 54(7): 763 71 of interventions that modify the aspects of parent infant interaction identified in this study. If intervention to optimise such interactions is to be undertaken, our findings suggest that it may be best to do this early before interaction patterns become embedded in emerging social atypicality. Supporting information Additional Supporting Information may be found in the online version of this article: Table S1 Brief description of the Manchester Assessment of Caregiver- Interaction (MACI) Table S2 Comparing the other concerns group with other 3-year outcome groups: 12-month mean (SD) AOSI and parent infant interaction ratings Table S3 Logistic regression models of the independent contributions of 6-month AOSI and domains of parent infant interaction to 3-year diagnostic classification Acknowledgements We are very grateful for the enormous contributions BASIS families have made towards this study. Thanks also to Dr Richard Emsley, University of Manchester for statistical advice. The research was supported by awards from the UK Economic and Social Research Council (RES-000-22-3525) to M.W. Wan and J. Green, and awards from UK Medical Research Council (G0701484), BASIS funding consortium led by Autistica (http://www.basisnetwork.org) and from Autism Speaks (1292) to M.H. Johnson. We are grateful to the anonymous reviews and the journal editorial process that helped us improve the article in revision. Correspondence Ming Wai Wan, Institute of Brain, Behaviour and Mental Health, Jean McFarlane Building, University of Manchester, Oxford Road, Manchester, UK; Email: m.w.wan@manchester.ac.uk Key points Increasing evidence suggests that studying early parent infant interaction in emergent ASD may be important for a more complete understanding of the development of the disorder. Interaction differences according to ASD/risk status were found at 6 months in parent interactive behaviours and infant liveliness, and at 12 months in parent interactive behaviours, infant attentiveness to parent, infant positive affect and dyadic mutuality in interaction. interactive behaviours and dyadic mutuality in interaction at 12 months, but not 6 months, predicted 3- year ASD outcome, independent of infant age and behavioural atypicality. Intervention efforts to optimise social functioning may need to start early in infancy before parent child interaction patterns become embedded in emerging social atypicality. References Baker, J.K., Messinger, D.S., Lyons, K.K., & Grantz, C.J. (2010). A pilot study of maternal sensitivity in the context of emergent ASD. Journal of ASD and Developmental Disorders, 40, 988 999. Bryson, S.E., Zwaigenbaum, L., McDermott, C., Rombough, V., & Brian, J. (2008). The ASD observation scale for infants: Scale development and reliability data. Journal of ASD and Developmental Disorders, 38, 731 738. Dawson, G. (2008). Early behavioural intervention, brain plasticity, and the prevention of autism spectrum disorder. Development and Psychopathology, 20, 775 803. Dennis, M., Francis, D.J., Cirino, P.T., Schachar, R., Barnes, M.A., & Fletcher, J.M. (2009). Why IQ is not a covariate in cognitive studies of neurodevelopmental disorders. Journal of the International Neuropsychology Society, 15, 331 343. Elsabbagh, M., Mercure, E., Hudry, K., Chandler, S., Pasco, G., Charman, T., & the BASIS Team. (2012). neural sensitivity to dynamic eye gaze is associated with later emerging ASD. Current Biology, 22, 338 342. Feldman, R., & Greenbaum, C.W. (1997). Affect regulation and synchrony in mother infant play as precursors to the development of symbolic competence. Mental Health Journal, 18, 4 23. Goodman, R., Ford, T., Richards, H., Gatward, R., & Meltzer, H. (2000). The Development and Well-Being Assessment: Description and initial validation of an integrated assessment of child and adolescent psychopathology. Journal of Child Psychology and Psychiatry, 41, 645 655. Key, A.P.F., & Stone, W.L. (2012, online). Same but different: 9-month-old infants at average and high risk for autism look at the same facial features but process them using different brain mechanisms. Autism Research 5, 253 266. Lord, C., Rutter, M., & Couteur, A. (1994). Autism diagnostic interview-revised: a revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders, 24, 659 685. Lord, C., Rutter, M., Goode, S., Heemsbergen, J., Jordan, H., Mawhood, L., & Schopler, E. (1989). ASD diagnostic observation schedule: a standardized observation of communicative and social behavior. Journal of ASD and Developmental Disorders, 19, 185 212. Losh, M., Childress, D., Lam, K., & Piven, J. (2008). Defining key features of the broad ASD phenotype: a comparison across parents of multiple- and single-incidence ASD families. American Journal of Medical Genetics B, 147, 424 433. Meirsschaut, M., Warreyn, P., & Roeyers, H. (2011). What is the impact of autism on mother-child interactions within families with a child with autism spectrum disorder? Autism Research, 4, 358 367. Mullen, E.M. (1995). Mullen scales of early learning. Circle Pines, MN: American Guidance Service. NICHD Early Child Care Research Network. (2001). Child care and children s peer interaction at 24 and 36 months: The NICHD study of early child care. Child Development, 72, 1478 1500. Ozonoff, S., Young, G.S., Carter, A., Messinger, D., Yirmiya, N., Zwaigenbaum, L., & Stone, W.L. (2011, online). Recurrence

doi:10.1111/jcpp.12032 At-risk infant interaction and autism 771 risk for autism spectrum disorders: A baby siblings research consortium study. Pediatrics, 128, e488 e495. Rozga, A., Hutman, T., Young, G.S., Rogers, S.J., Ozonoff, S., Dapretto, M., & Sigman, M. (2011). Behavioral profiles of affected and unaffected siblings of children with ASD: Contribution of measures of mother-infant interaction and nonverbal communication. Journal of ASD and Developmental Disorders, 41, 287 301. Rutter, M., Bailey, A., & Lord, C. (2003). Social communication questionnaire. L.A.: WPS. Saint-Georges, C., Cassesl, R.S., Cohen, D., Chetouani, M., Laznik, M.-C., Maestro, S., & Muratori, F. (2010). What studies of family home movies can teach us about autistic infants: A literature review. Research in Autism Spectrum Disorders, 4, 355 366. Saint-Georges, C., Mahdhaoui, A., Chetouani, M., Cassel, R.S., Laznik, M..C., Apicella, F., & Cohen, D. (2011). Do parents recognize autistic deviant behavior long before diagnosis? Taking into account interaction using computational methods. PLoS One, 6, 1 13. Sameroff, A.J. (Ed.), (2009). The transactional model of development: How children and contexts shape each other. Washington, DC: American Psychological Association. Sameroff, A.J., & Fiese, B.H. (1990). Transactional regulation and early intervention. In S.J. Meisels & J.P. Shonkoff (Eds.), Handbook of early childhood intervention (pp. 119 149). New York: Cambridge University Press. Schwichtenberg, A.J., Young, G.S., Sigman, M., Hutman, T., & Ozonoff, S. (2010). Can family affectedness inform infant sibling outcomes of autism spectrum disorders? Journal of Child Psychology and Psychiatry, 51, 1021 1030. Wallace, K.S., Rogers, S.J. (2010). Intervening in infancy: Implications for autism spectrum disorders. Journal of Child Psychology and Psychiatry, 51, 1300 1320. Wan, M.W., Green, J., Elsabbagh, M., Johnson, M., Charman, T., Plummer, F., & The BASIS Team. (2012). Parent-infant interaction in infant siblings at risk of ASD. Research in Developmental Disabilities, 33, 924 932. World Health Organisation. (1993). The International Classification of Diseases: classification of mental and behavioural disorders. Diagnostic criteria for research. (ICD-10) (10th edn). Geneva: WHO. Virkud, Y.V., Todd, R.D., Abbacchi, A.M., Zhang, Y., & Constantino, J.N. (2009). Familial aggregation of quantitative autistic traits in multiplex versus simplex autism. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 150B, 328 334. Yirmiya, N., & Charman, T. (2010). The prodrome of autism: Early behavioral and biological signs, regression, peri- and post-natal development and genetics. Journal of Child Psychology and Psychiatry, 51, 432 458. Yirmiya, N., Gamliel, I., Pilowsky, T., Feldman, R., Baron- Cohen, S., & Sigman, M. (2006). The development of siblings of children with ASD at 4 and 14 months: Social engagement, communication, and cognition. Journal of Child Psychology and Psychiatry, 47, 511 523. Zwaigenbaum, L., Bryson, S., Rogers, T., Roberts, W.B.J., & Szatmari, P. (2005). Behavioural manifestations of autism in the first year of life. International Journal of Developmental Neuroscience, 23, 143 152. Accepted for publication: 12 October 2012 Published online: 11 December 2012