Autism Spectrum Disorder in Children Adopted After Early Care Breakdown

Findings from longitudinal studies of children adopted internationally following severe early global deprivation in orphanages have demonstrated a significant impact of adverse environmental experience on later social development, including a pattern of ‘quasi-autism’ (Q–A), found for instance in 6.3 % of 165 children at 6 years in the England-Romanian Adoptees (ERA) study, with a similar additional number of children displaying partial features (Rutter et al. 1999). Key features of the described syndrome were difficulties in forming selective friendships, impaired social reciprocity, eye contact, social gesture and a lack of reciprocal social exchange, sensory preoccupations and intense circumscribed interests (as opposed to the stereotyped behaviours such as rocking previously associated with institutional rearing). The profile differed from ‘classical’ autism in an increased amount of social approach and spontaneous communication, equal sex ratio, non-elevated head circumference and (as it proved on further follow-up) a developmental course rather unlike classical autism. Thus in ERA at 11 years the Q–A pattern persisted in 75 % of children, but in others had become modified to resemble the unusual social behaviour known as Disinhibited Attachment Disorder (DAD), observed separately in 39.1 % of the cohort at 11 years (Rutter et al. 2007; Kumsta et al. 2010). At 18–20 years, 15 young people still met criteria for Q–A (compared to 16 at 6 years), the majority of these cases still showed Q–A characteristics such as unusual obsessional behaviours and circumscribed interests (Kreppner et al. 2010), but there had often been a lessening in social impairment, and frequent overlap with criteria for DAD (Kumsta et al. 2010). Severe early institutional deprivation was thus associated with a striking autism-like syndrome in early childhood in up to 13 % of the sample - substantially higher than would be expected from population-based prevalence rates for ASD, which are consistently around 1 % across global studies (Elsabbagh et al. 2012). The developmental course of this Q–A tends to modify towards the broader autism phenotype, although core social impairments usually persist; presence of Q–A in adolescence is associated with raised rates of other psychopathology, functional impairment and increased service use (Kumsta et al. 2010; Kreppner et al. 2010).

This identification of Q–A was in a very particular population group and a key question follows as to whether similar impairments might also occur in children who had been adopted after severe early social neglect or maltreatment within families (since high income societies have largely abolished early institutional care). A national survey of 1253 children aged 5–17 years in local authority care in Britain found a prevalence of caregiver-reported autism-like symptoms of 2.6 %, but there was no in-person phenotypic ascertainment (Ford et al. 2007). ‘Autistic-like detachment’ was described in 26 % of 70 children at care-entry between 5 and 12 years (Dimigen et al. 1999) but this had not been assessed in relation to defined Q–A or ASD. More recently, 46 % of 35 children (5–8 years) selected for Reactive Attachment Disorder (RAD) after early maltreatment showed autism symptoms on standard caregiver interview (Sadiq et al. 2012). Such studies suggest the possibility of increased rates of ASD in high-risk children in out-of-home care, but lack direct and detailed phenotypic ascertainment. Further, they beg the aetiological question of the origin of any ASD found. In the ERA study, a primary effect of the institutional environment was supported by a dose–response relationship between length of post-natal institutional exposure and Q–A outcomes, along with available data suggesting that prior familial and prenatal biological risks were unlikely to account for the symptoms (Kumsta et al. 2010). However, in the context of familial neglect and maltreatment equivalent assumptions cannot be made: as well as the potential direct effect of severe neglect or maltreatment itself there may be a convergence of other pre-, peri-, and post-natal risks associated with a maltreating environment, such as familial genetic risk, prenatal exposure to alcohol, drugs, infection, prematurity, poor nutrition, or major psychological stress, all themselves recently linked with potential neurodevelopmental impacts (Miles et al. 2003; Ronald et al. 2011; Rai et al. 2013; Bromley et al. 2013; O’Connor et al. 2014).

Against this background, we designed to our knowledge the first systematic ascertainment of possible ASD in children adopted after early care breakdown, a group associated with severe early neglect or maltreatment (Rees and Selwyn 2009) but likely also to have suffered other pre-natal risk exposure. Phenotypic ascertainment used a two-phase design, with initial standardised screen followed by in-person phenotypic ascertainment using gold standard assessment instruments for ASD. The study was also designed to address as completely as possible, potential familial, prenatal or postnatal risk exposures that might be associated with any ASD identified.

This two phase ascertainment study of children in adoption after early disrupted care, neglect or maltreatment used standardised measures and demonstrated a strikingly high incidence of ASD phenotype in 11 %, with a further 18.5 % showing partial features. The children showing ASD are not just presenting with impairment of social reciprocity and empathy, they show the full range of characteristic DSM5 ASD symptoms including repetitive behaviours, unusual interests and stereotyped behaviours (although generally a lack of atypical early language). The pattern of findings on phenotypic ascertainment, confirmed by clinical assessment, suggest a core group of three cases with a family history of autism, learning disability or mental illness, who were removed at or very soon after birth and thus exposed to little or no postnatal environmental family risk. These cases seem characteristic of a familial idiopathic autism or other pre-natal risk, which has clustered in families whose parents have proved unable to manage a child who has come into early care. A further group of three PDD-NOS cases do not show birth-family history of autism but do report birth family learning difficulty and/or mental health problems with reported exposure to pre-and/or post-natal adversity; notably, high levels of parental mental illness, small head circumference, post-natal neglect or physical abuse. All PDD-NOS cases showed a full range of ASD symptomatology but in less severe form; all but one showed a full or partial ASD syndrome on clinical assessment. Of the ten cases showing ‘Broad ASD’, clinical assessment on four reported ASD or partial traits (one of these with comorbid DAD); two showed DAD without ASD traits; three showed ADHD (one comorbid with ASD); and others a mixture of non-specific attachment and sensory processing features.

The approach to in-person ASD phenotype ascertainment used in this study allows comparison with results from the equivalent assessment approach taken in the Romanian adoptees study (Rutter et al. 1999). The ASD syndrome profile shows a similar range in both studies, notably including repetitive, stereotyped behaviours and restricted interests as well as social impairment; and similar features such as the equal gender ratio and lack of atypical head size. However, as noted in the introduction, the context of the two cohorts differs. Whereas the Romanian adoptees the focus of difficulty was on the post-natal social deprivation also seen in our sample, here we find in addition pre-natal and family risk exposures, which are increasingly typical of children coming into adopted care. Q–A in the ERA cohort showed a unique developmental progress in development that was different to autism, longitudinal follow-up in this study is underway and will be important to understand the developmental course of our findings. Our study is unique to our knowledge in using standardised in-person assessment in this context, but its findings are consistent with other research on children post adoption. Thus in-depth study of 35 families from a recent survey of 390 adoptive parents in England (Selwyn et al. 2014) reported independent autism diagnosis in 11 %; earlier accounts of autistic-like detachment in 26 % of 70 children at entry into care (Dimigen et al. 1999) and impaired theory of mind in children after maltreatment (Cicchetti et al. 2003; Kay and Green 2015).

In addition to ASD, we report high levels of co-morbidity across all the screen-positive group (see Table 2) and this is an important feature of this clinical area. However, the rates of comorbidity do not vary between autism, PDDNOS or Broad Autism categories, and suggests that our CPEA-based research diagnoses of autism are specific and not misdiagnoses confounded by concurrent other disorder. Between screen positive and screen-negative cases levels of emotional disorder and ADHD do differ; this may reflect common-rater biases operating in parent-report based measures and emphasizes the value of in-person assessment of ASD.

Regarding limitations: we consider the findings in this sample as striking and important in their own right but cannot be sure yet quite how typical they are of UK adoption as a whole. The families self-referred in response to an invitation worded to be non-specific without mention of autism or any other specific hypothesis. The majority of children referred had suffered significant neglect and maltreatment and in this they are very typical of the general pattern now current within UK adoption where 72 % of children with adoption orders first become looked after due to maltreatment at an average age of 1.2 years, and roughly two thirds are adopted before the age of 4 years (Selwyn et al. 2014). We have no evidence therefore of significant sampling bias affecting our results, although this cannot be excluded. Confirmatory survey data is planned to establish prevalence rates applicable to the wider population of UK adopted children. The method of ascertainment of prior risk exposures through adoptive parent report was consistent across the study, but it is clearly possible that, despite our best efforts, the developmental interview data under-represents or biases the actual extent of risk exposure in these children. The practicality of accessing confirmatory medical or social care records was very variable, due to the lack of contact with social services and birth families, the need for birth family consent for records access and the time elapsed since adoption; it was considered that using the records in some cases but not others would be misleading and result itself in differential assessment biases. In the current environment of UK adoption (typical of other high income countries), there was no opportunity to design a comparison group of children adopted without early adversity. For instance, the group of children adopted at birth (thus avoiding postnatal adversity) are almost universally adopted because of overwhelming concern about familial or prenatal risk exposures - and our data reflects this. Detailed phenotyping was untaken on screen positive cases only; resources precluded assessing screen negative cases also. However, our study was focused on a conservative ascertainment of ASD rather than testing the specificity of the screen measure, and we think the risk of not having identified true ASD cases in the sample is low. Our identification of ASD using CPEA criteria was chosen as the best available systematic way to integrate the ADI and ADOS research assessments into a classification scheme; it does not replicate a full clinical ascertainment but has shown added value in presenting a standardized quantifiable approach to classification which can avoid the vagaries of clinical diagnostic usage (Lord et al. 2012a, b).

These results are suggestive but do not yet allow specific inferences for developmental theory. Our data suggests that there are likely be a variety of convergent familial, pre-and postnatal risks responsible for the high rates of ASD symptoms seen here but precisely which and in how they are interacting in most cases cannot be clearly determined from this retrospective study. The majority of children had significant exposure to severe neglect and maltreatment in their birth families but there is no simple evidence of a “dosage” effect of length of postnatal exposure in relation to later outcome, as was found within the ERA study (Kumsta et al. 2010). Further hypothesis-driven research designs are indicated, ideally including prospective ascertainment of at risk groups in the prenatal period.

The clinical implications of these results are however immediate and significant, since identification of ASD impairments has specific implications for family understanding, style of intervention, and educational planning. Anecdotally, adopting families across our study often reported feeling isolated, with a lack of appropriate services and unsure as to whether to ascribe their child’s difficulties to attachment or emotional disorder, stubbornness, rejection, or developmental difficulty; reports that echo recent large-scale formal surveys of adopting parents (Pennington 2012; Selwyn et al. 2014). Services themselves are faced with adjustment to the increasing complexity in risk background of children coming into adoption; a common response to a child’s difficulty in engagement with an adoptive placement or destructive or disruptive behaviour has been to ascribe a psychological model of attachment disruption, emotional inhibition or post-traumatic disturbance. Clearly all of these phenomena may be significant in children with such pasts, but the presence of neurodevelopmental disorder such as ASD is an important alternative explanation for the child’s difficulties, with different implications. Informed assessment that can characterise developmental disorder as well as psychological disturbance in these complex cases is essential and will lead to much more efficient and targeted management. It is also possible that these social impairments, if persistent, could contribute to the very poor functional outcomes sometimes reported after adoption; the extent of functional impairment associated with quasi autism or attachment disorder in the ERA cohort emphasise this point. Understanding of the neurodevelopmental consequences of the combination of early biological and environmental risk exposure should result in a paradigm shift; opening up new ways of understanding and managing the problems that often arise now for children and families post-adoption.