Inattention/Overactivity Following Early Severe Institutional Deprivation: Presentation and Associations in Early Adolescence

Transcription

1 J Abnorm Child Psychol (2008) 36: DOI /s Inattention/Overactivity Following Early Severe Institutional Deprivation: Presentation and Associations in Early Adolescence Suzanne E. Stevens & Edmund J. S. Sonuga-Barke & Jana M. Kreppner & Celia Beckett & Jenny Castle & Emma Colvert & Christine Groothues & Amanda Hawkins & Michael Rutter Published online: 27 October 2007 # Springer Science + Business Media, LLC 2007 Abstract The current study examined the persistence and phenotypic presentation of inattention/overactivity (I/O) into early adolescence, in a sample of institution reared (IR) children adopted from Romania before the age of 43 months. Total sample comprised 144 IR and 21 non-ir Romanian adoptees, and a comparison group of 52 within-uk adoptees, assessed at ages 6 and 11 years. I/O was rated using Rutter Scales completed by parents and teachers. I/O continued to be strongly associated with institutional deprivation, with continuities between ages 6 and 11 outcomes. There were higher rates of deprivation-related I/O in boys than girls, and I/O was strongly associated with conduct problems, disinhibited attachment and executive function but not IQ more generally, independently of gender. Deprivation-related I/O shares many common features with ADHD, despite its different etiology and S. E. Stevens : E. J. S. Sonuga-Barke Developmental Brain-Behaviour Unit, School of Psychology, University of Southampton, Southampton, UK S. E. Stevens : E. J. S. Sonuga-Barke : J. M. Kreppner : C. Beckett : J. Castle : E. Colvert : C. Groothues : A. Hawkins : M. Rutter MRC Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, King s College London, London, UK E. J. S. Sonuga-Barke Child Study Center, New York University, New York, NY, USA S. E. Stevens (*) SGDP Centre, Box PO.80, Institute of Psychiatry, King s College London, De Crespigny Park, London SE5 8AF, UK s.stevens@iop.kcl.ac.uk putative developmental mechanisms. I/O is a persistent domain of impairment following early institutional deprivation of 6 months or more, suggesting there may be a possible pathway to impairment through some form of neuro-developmental programming during critical periods of early development. Keywords Inattention/overactivity. Early deprivation. Romanian institutional rearing. International adoption Introduction Inattention, overactivity (I/O) and impulsiveness, the cluster of behavioral features that form the diagnostic core of Attention Deficit/Hyperactivity Disorder (ADHD), are amongst the most common clinical indications amongst children raised during their early years in institutions (e.g. Goldfarb 1945; Tizard and Hodges 1978; Fisher et al. 1997; Roy et al. 2000, 2004). Probably some of the most compelling evidence for I/O as a specific outcome of institutional care has come from the English and Romanian Adoptees (ERA) study (Kreppner et al. 2001; Rutter et al. 2001). ERA, like a number of recent projects (Gunnar and van Dulmen 2007; Marcovitch et al. 1995; Maclean 2003), examined the putative causal role that early adverse experiences associated with institutional deprivation play in determining developmental outcome (see Rutter and ERA Study Team 1998; O Connor et al. 2000). The ongoing study comprises a large representative sample of Romanian children who were raised in the severely deprived conditions of the state institutions of the Ceauşescu regime in Romania at the end of the 1980s before they were adopted by families living in the UK. The grave nature of the situation in

2 386 J Abnorm Child Psychol (2008) 36: which the children were reared was apparent in the marked developmental delay and poor physical state of the children at the time of entry to the UK. Many children demonstrated catch-up in the physical and intellectual domains by the time they were 4 6 years of age although residual deficits persisted in a significant minority of children (Rutter and ERA Study Team 1998; O Connor et al. 2000) and psychological dysfunction and psychiatric morbidity were common within the sample. Deficits were surprisingly specific and unusual in pattern and found to be associated with duration of time spent in the depriving conditions of the Romanian institutions (Rutter et al. 2000, 2001). In this regard, four domains of impairment were implicated: cognitive impairment, disinhibited attachment, quasi-autism and, most significantly for the current paper, I/O. Considerable individual continuity in normality and impairment was found between ages 6 and 11 years, with pervasive and persistent impairment across multiple domains of functioning in those children who had experienced at least 6 months_ institutional rearing (Kreppner et al. 2007). There were also specific, persistent adverse effects into early adolescence on cognitive impairment (Beckett et al. 2006) and disinhibited attachment (Rutter et al. 2007a) in this group who had experienced extended deprivation, and clinically significant rates of quasi-autistic patterns (Rutter et al. 2007b). Additionally, there was a significant increase in emotional difficulties by age 11 that was strongly associated with previous impairment in the deprivation specific problem areas identified at age 6 (Colvert et al. in press). A number of features of the study design have helped to facilitate the interpretation of these effects and served to strengthen the conviction that they were specific to early global institutional deprivation. These include: a relatively large sample (N>150), the availability of detailed information about the timing and extent of institutional care for each child and their age at adoption, the measurement of longitudinal outcomes at multiple follow ups (ages 4, 6, and 11 years with an age 15 phase underway) across a broad range of assessment domains (intellectual, behavioral and social), and a relative homogeneity in the sample with regard to likely factors responsible for entry to institutions and its timing. Most children entered the institutions within the first two weeks of life and, although we do not have any systematic information on the reasons for placement, evidence from surveys conducted at the time (Children s Health Care Collaborative Study Group 1992) and the early age at which the children were admitted indicate that this was due to extreme poverty and social exclusion. Moreover, there was an absence of any formal fostering system within Romania and, as far as is known, no children were adopted from the institutions prior to Thus the subsequent timing of adoption out of institutions was largely determined by political, rather than individual selection, factors brought about by the fall of the Ceauşescu regime in 1989, following which adoption became possible. The study also included a comparison group of children born and adopted in the UK before the age of six months (N>50), and a sampling approach that ensured three large, similarly-sized, groups including Romanian children entering the UK and placed with their eventual adoptive families during different periods in the first 3 1/2years of life (i.e., entering the UK before the age of 6 months, between 6 and 24 months and 24 months or over). This final element allowed the relationship between duration of deprivation (used as a proxy for dose of deprivation) and outcome to be established. At ages 4 and 6, outcomes, including I/O, generally demonstrated a linear-like dose response relationship with duration of deprivation. In addition, there was a striking level of heterogeneity in outcomes. Many children, even in the group with highest dose of deprivation, were well adjusted and unimpaired. All in all, despite this heterogeneity, the ERA data support the idea that I/O at age 6 years was a specific outcome of early institutional deprivation for some, although not all, children. The current paper had three aims: first, to examine the persistence and continuity of the I/O pattern, and its specific links with duration of deprivation from childhood into early adolescence. As noted above, I/O was isolated as a specific area of deficit at ages 4 and 6 years and was strongly associated with duration of time spent in the extremely deprived conditions of the Romanian institutions. The current paper extends previous analyses in two ways: firstly, by looking at persistence to age 11 and, secondly, by including the most deprived group (aged 24 months or over at adoption) in the analysis of change over time. Persistence has not previously been assessed in this group as they were too old to be included in the 4 year old phase of the study. By age 11 all the children had spent at least 7 1/2years in their adoptive homes. If the effects seen at age 6 persisted despite this, the possibility that I/O represents a transient behavioral reaction to deprivation becomes less likely and the possibility that it reflects an early established and fundamental brain-mediated effect becomes more likely. Second, we aimed to characterize what is specific to the deprivation-related I/O phenotype in order to contrast it with I/O in the general non-deprived population. It has often been supposed that deprivation-related I/O is a qualitatively distinct entity from I/O in the normal population owing to different putative causal mechanisms, for instance as is seen in children on the ADHD spectrum (Roy et al. 2004). By definition it has a different etiology: nondeprivation-related I/O is highly heritable and strongly related to susceptibility genes. Where environmental factors are implicated these are often related to pre- and perinatal adversity rather than post-

3 J Abnorm Child Psychol (2008) 36: natal social and interpersonal factors, although the potential role of prenatal and genetic factors in deprivation-related I/O cannot be ruled out (Taylor and Warner-Rogers 2005). One might therefore expect deprivation and nondeprivationrelated I/O to be mediated and moderated by different factors and have different patterns of associations. In order to explore whether this is true we examined the association between deprivation-related I/O and four factors consistently shown to be associated with nondeprivation-related I/O; (i) its male preponderance, (ii) its cross-sectional and longitudinal link with conduct problems, and its association with (iii) low IQ and (iv) executive function deficits. Gender imbalance Although the picture is far from clear regarding the causes of gender differences in ADHD the discrepancy in prevalence rates is undisputed, with ratios of girls to boys reported to be between 1:2 and 1:9 (Youth in Mind 2001; Heptinstall and Taylor 2002; Biederman et al. 2002). While there may be a degree of rater bias this cannot explain the phenomenon fully (Maniadaki et al. 2005). Girls may be more resilient in relation to risks for the development of ADHD and differences in cognitive impairment, comorbid behavior problems and some discrepancies in symptomatology have been noted (Heptinstall and Taylor 2002). However, in our institution-reared (IR) sample at age 6 there was a fairly even distribution of deprivation-related I/O across boys and girls (Kreppner et al. 2001). One possible reason for this is that early institutional deprivation is a particularly potent risk factor for female I/O that combines with other risks in a way that pushes certain girls over their risk threshold for the expression of the condition. Conduct problems Nondeprivation-related I/O and conduct problems often co-occur. Studies of clinic and populationderived samples of children and adolescents have found a high rate of ADHD cases comorbid with conduct disorder (CD) or oppositional defiant disorder (ODD): in the region of 40 90% (Jensen et al. 1997). This pattern of comorbidity is a common and pervasive long-term adverse outcome with strong homotypic continuity over time (Willcutt et al. 1999; Burke et al. 2005). Although it has been suggested that ADHD comorbid with CD may share a common set of genetic risk factors and may represent a genetically more severe type of ADHD, research has supported the distinction of these two domains of dysfunction (Thapar et al. 2001). Developmental studies have suggested that the presence of early ADHD predicts the occurrence of ODD and subsequent CD, but ODD does not predict the later emergence of ADHD (Burke et al. 2005; Taylor et al. 1996). In addition to genetic influences, it is plausible that the similar set of environmental risk factors such as pre- and perinatal adversity, and psychosocial/family risk, associated with both ADHD and conduct/oppositional problems could help to account for the progression from one condition to the other (Thapar et al. 2006). The findings from the ERA study suggested that at age 6 and age 11 conduct problems were not a specific outcome of the deprivation experience (i.e., related to dose of deprivation) (Colvert et al. in press) but the relation to I/O in deprived samples has not been investigated. There has been mixed evidence from other samples of post institutionalized children in relation to whether increases in levels of conduct and oppositional problems are observed (Gunnar and van Dulmen 2007). Low IQ The negative association between IQ and nondeprivation-related I/O symptoms has been consistently reported. There is typically a correlation of around 0.3 between ADHD symptom scores or diagnosis and IQ (Kuntsi et al. 2004) representing a deficit of between 9 and 13 IQ points (Rucklidge and Tannock 2001; Crosbie and Schachar 2001) compared with normal controls. The nature of this association is open to several interpretations (see Goodman et al. 1995). Goodman et al. theorized that hyperactive behavior may interfere with learning success or performance on IQ tests or perhaps that low IQ increases the risk for hyperactivity via its association with reduced self esteem. Low IQ and I/O could also be markers of some common underlying risk factor or factors such as variations in brain development, individual genetic makeup or shared environmental adversity (Goodman et al. 1995; Kuntsi et al. 2004). Executive dysfunction The dominant model of the psychopathophysiology of ADHD has focused on the role of executive dysfunctions involving multifaceted deficits in neurocognitive processes, such as working memory, response inhibition and interference control, which maintain and manage appropriate information and problem solving sets in order to achieve a future cognitive goal (Castellanos et al. 2006). A recent meta-analysis by Willcutt et al. (2005) demonstrated significant case-control differences, with medium effect sizes (d= ) in several key domains: response inhibition, vigilance, spatial working memory and some planning tasks, which were independent of IQ, academic attainment or comorbid disorders. Our third aim was more exploratory and related to one of the most obvious areas where deprivation and nondeprivation-related I/O might differ: the association with reactive attachment disorder of the disinhibited subtype, a common outcome in studies of institutionalized children (Zeanah et al. 2005; Chisholm 1998; Rutter et al. 2007a; Roy et al. 2004). This consisted of an unusual pattern of disinhibited approach to strangers and was noted by parents and observed by study investigators. Study of the relationship between I/O and attachment in non-institutionalized

4 388 J Abnorm Child Psychol (2008) 36: samples has been limited. However, attachment theory holds that a secure and responsive early parent child relationship is an integral part of the development of effective selfregulation in the child and self-regulation is linked to impulse control, perseverance and inhibition, which make up important features of the nondeprivation-related I/O and ADHD phenotype. Most studies have focused on insecure attachment with parents, rather than disinhibition with strangers, and are mainly based on small clinical case studies (Stiefel 1997; Clarke et al. 2002). Nevertheless, in combination with the striking pattern of disinhibited attachment observed in our sample and pattern of overlap noted by Kreppner et al. (2001), these studies highlight this as an important area of investigation when considering the phenotypic characteristics of I/O in adolescence. The aim of the current paper was to explore deprivationrelated I/O in early adolescence by asking the following questions: 1. Does the risk for I/O associated with severe early institutional deprivation persist to age 11 years? 2. Is there individual continuity in I/O behavior? 3. Is deprivation-related I/O phenotypically similar to ADHD/hyperactivity in terms of: a. The gender imbalance in prevalence rates? b. The comorbidity and developmental association with conduct problems? c. The association with low IQ? d. The association with executive function? 4. Is there overlap between I/O and disinhibited attachment? Materials and Methods Sample The sample of 165 Romanian children adopted before the age of 43 months was drawn from 324 children adopted into UK families between February 1990 and September The adoptive parents did not always know the ethnicity of the children they adopted so the collection of systematic data on ethnicity was not possible. However, in Romania as a whole the population is over 90% Romanian, defined by a common language, and includes a substantial minority of Roma people (estimates vary between 5 10% of the population) and also people from neighbouring countries e.g. Hungary and Russia. At the time of entry to the UK only a small minority of children possessed even the most basic language skills. Of the children aged 18 months or over (the age by which in a normal population the vast majority of children would be attempting to reproduce words) only 13 out of 57 were using 3 recognizable words and none had even minimal fluency in spoken Romanian language, despite the age range of the children reaching 3 1/2years. Language development throughout the sample was tested at age 6 and all assessments were carried out in English (Croft et al. 2007). The sample was balanced for gender, selected at random from within two age bands (<6 months; 6 to <24 months) and stratified according to age at entry to the UK; all the families of the children of 24 months of age were selected, as there were fewer children in this age band. Of the 165 children in the Romanian sample, 144 were raised, usually from soon after birth, in the extremely depriving conditions of the state institutions. The remaining 21 Romanian children were adopted from family settings and their ages at adoption were spread throughout the 3 main ages bands. This group of non-institutionalized children provided a useful comparison in that they experienced the general hardship and poverty suffered by underprivileged Romanian families at the time but they were not subject to the experience of institutional rearing and its associated risks. A comparison sample of 52 within country (UK) adoptees aged below 6 months when adopted was selected and obtained through voluntary and local authority adoption agencies. The comparison sample was chosen in order to control for the experience of adoption and of being brought up, post adoption, in an above average rearing environment, but to vary in terms of the experience of early severe psychological and nutritional deprivation. For a more comprehensive description of the sample, including sampling and assessment strategy, post adoption environment and the physical and developmental condition of the children at the time of adoption, see O Connor et al. (2000). For 210 of the 217 (97%) children in the study data on I/O at age 11 had been collected from either parent (n=199, 92%) or teacher (n=188, 87%) reports. The seven children with missing data were all from the Romanian sample and were split evenly across the 3 main age bands (<6: n=2; 6 to <24: n=2; 24: n=3) with slightly more girls than boys with data missing (girls: n=5; boys: n=2). Measures The full sample was assessed at ages 6 and 11 years using a combination of standardized tests, standardized investigatorbased interviews, qualitative interviews, questionnaires and observations. In-depth interviews were carried out with adoptive parents and questionnaires on children s behavior were completed by parents and teachers. Behavioral assessment of I/O and conduct problems The Revised Rutter Parent and Teacher Scales for school-age children (Elander and Rutter 1996) with supplementary ques-

5 J Abnorm Child Psychol (2008) 36: tions from Behar and Stringfield (Behar 1977; Hogg et al. 1997) were administered at ages 6 and 11years. The questionnaires were completed at both assessment time-points by mothers, fathers and teachers. At the age 11 time-point the questionnaires were completed by each child s primary school main class teacher, i.e. before they moved to secondary education. The scales comprised sets of items describing different behaviors; each item, or statement, is scored on a scale of 0 2: 0 for doesn t apply, 1 for applies somewhat, 2 for certainly applies. The individual items used to assess these domains are listed in Appendix. Parent and teacher composite scores were created for both the I/O and conduct problems subscales. A combined parent score was calculated by taking the mean mother and mean father scores across the questionnaire items and then calculating an average of the two. In order to maximize the sample size, children who had obtained ratings from only one parent were also included. The correlation between mother and father reports of I/O behaviors on the Rutter Scales was high, with medium to high effect sizes found (Age 6: r(175)=0.74; p<0.001; r 2 =0.54. Age 11: r(174)= 0.78; p<0.001; r 2 =0.61). Teacher scores were calculated by taking the mean score across the items for each behavioral domain. In order to examine markedly abnormal behavior and to compare with prevalence rates in the population, cut-offs were calculated by transforming the continuous outcome measure of I/O into categorical data. There were no established cut-off criteria for the Rutter subscales and therefore the following strategy was developed based on the procedure used for assessing behavior rated on the Strengths and Difficulties Questionnaire (SDQ; Goodman 1997). This also allowed us to compare rates of problem behavior with those from a population sample using normative data on the SDQ. The SDQ comprises sets of behavioral description items based largely on the Rutter Scales that are scored on a similar 3 point scale: 0 for not true, 1 for somewhat true and 2 for certainly true (see Appendix). As reported by Goodman (1997), the Rutter Scales and the SDQ are very highly correlated, r(346) =0.88, 0.88, 0.82 and r(185) = 0.92, 0.91, 0.90 in terms of total difficulties, hyperactivity and conduct problems scales according to parent and teacher ratings, respectively. The equivalence of the two scales in terms of correlation, behavior scale items and rating structure justified the application of a cut-off from one scale being applied to the other. The cut-off was calculated for the Rutter Scales according to the procedure for determining behavior in the abnormal range as outlined on the official SDQ-info website (Youth in Mind, pdf; Goodman 1997). For the hyperactivity subscale on the SDQ a score of 7 or above on the summed composite (score of 0, 1 or 2 per item on a five question subscale; making a possible total score of 10) was considered in the abnormal range. The abnormal cut-off was transformed for the Rutter Scales by taking the lower limit of the abnormal banding on the SDQ and dividing by the number of items on the SDQ hyperactivity scale to obtain an average score per item, at or above which would be considered abnormal: 7 (lower limit) 5 (items)=1.4. This cut-off was then applied to the mean scores of parents and teachers on the Rutter Scales. For the analyses of the patterns of association between deprivation-related I/O in the Romanian-IR risk sample aged 6 42months at adoption and the various phenotypic features the subsample was split according to their level of I/O impairment. There were three I/O impairment groups: firstly, those children who were below cut-off according to both parents and teacher were classed as being in the normal range; secondly, the children rated as above cutoff according to parents or teachers were classified as having situational I/O; lastly, those children rated above cut-off according to both parents and teachers were designated as pervasive. To address questions about continuity of I/O impairment or normality the situational and pervasive I/O categories were collapsed to create a dichotomous split between those in the abnormal range according to parent and/or teacher and those in the normal range according to both informants. To compare prevalence rates in early adolescence of abnormal I/O behavior in the Romanian IR risk sample according to gender with the discrepancy seen in ADHD in the population it was necessary to restrict the analysis to those who showed I/O impairment according to parent and/ or teacher reports at age 6 and age 11, as ADHD diagnostic criteria require an onset before the age of 7years. This, however, did restrict the sample size as individuals had to have data at all four data-points in order to be included in this analysis (parent and teacher report at 6 and 11years). This applies also to the analyses of continuity and phenotypic features described above. Moreover, gender discrepancy was investigated by comparing prevalence rates in the Romanian IR risk sample with those found in the normal population. Normative data from a large representative British survey of child and adolescent mental health, which used the SDQ questionnaire, was exploited (Department of Health & Office for National Statistics: Meltzer et al. 2000). The sample included 10,438 individuals aged between 5 and 15 years. Complete SDQ information was obtained from 10,298 parents (99% of sample), 8,208 teachers (79% of sample) and 4, year olds (93% of this age band) (Youth in Mind, Assessment of cognitive and executive function Cognitive functioning was assessed using a short form of the Wechsler

6 390 J Abnorm Child Psychol (2008) 36: Intelligence Scales for Children (WISC III UK ; Wechsler 1992). Four subscales of the WISC were included in the battery: two measuring verbal abilities vocabulary and similarities; and two measuring performance abilities block design and object assembly. These four subtests were selected to provide a good estimate of full scale IQ (reliability coefficient=0.94, Sattler 2002) and were prorated to form a full scale IQ score. There were three Romanian IR children who have been excluded from the analysis of association between I/O, IQ and executive function as the severity of their cognitive impairment was of such a degree that these aspects of the assessment battery were not suitable and therefore not administered (Beckett et al. 2006). The concept of executive function covers a broad range of cognitive processes. In this study two aspects of executive functioning at age 11 were tested: interference/inhibitory control and verbal working memory. The measures used to assess these abilities were the Stroop Color Word Interference Test (Stroop 1935) and the backwards digit span subtest on the WISC III uk (Wechsler 1992). The Stroop task assesses the speed and accuracy with which participants can name the contrastive color of the ink in which color words are written (e.g. the words red, green and blue could be written in blue, yellow and red ink, respectively). The task is designed to assess interference control, since the default response to seeing words on a card is to read the words rather than to name the color of the ink, and this response must be inhibited to complete the task correctly. The task was administered twice: firstly, the child was required to read the words on the card and ignore the color of the ink and, secondly, they had to name the color of the ink and ignore the words themselves (the inhibitory stage of the task). For both trials the number of errors committed was recorded and analyses used the total number of errors on the first trial subtracted from the total errors on the second trial as the dependent measure. The backwards digit span subtest of the WISC was used as a test of verbal working memory. Participants repeated a series of digits in the reverse order to which they were orally presented. The score used in the current paper is the total raw score of successfully completed trials. Assessment of disinhibited attachment This was measured using questions included in the interviews carried out with parents when the children were aged 11 years concerning essential components of disinhibited attachment behaviour: wandering off; too friendly with strangers/lack of differentiation; and physical contact/lack of understanding social boundaries/personal space (see Rutter et al. 2007a). These items were scored on a three point scale: 0 for no abnormality, 1 for probable problem and 2 for marked problem. An overall score ranging from 0 to 6 was calculated by summing across the three items. Results Does the Risk for I/O Associated with Institutional Deprivation Persist to Age 11Years? Kreppner et al. (2001) reported that prolonged duration of institutional deprivation constituted a significant risk factor for I/O at age 6. In order to establish whether, firstly, institutional deprivation continued to be a significant risk for I/O into early adolescence and, secondly, whether the dose response association between duration of deprivation and level of impairment continued to show a linear-like pattern, a within sample comparison of the mean levels of I/O behavior at 11 years was carried out across informants and adoptee groups. Table 1 presents the means, standard deviations and percentages in the abnormal range for I/O across sample group (IR group split according to age at entry to UK), gender and informant. Analysis of variance was used to compare the UK comparison, Romanian IR (pooled <6, 6 to <24 and 24 month groups) and Romanian non-ir. There was a significant difference in I/O at age 11 according to both parent (F(2, 196) =4.90; p<0.01) and teacher reports (F(2, 185) =6.68; p<0.01). Post hoc comparisons (Tukey s test) found significantly higher levels of I/O in the Romanian IR sample (parent: M=0.70, SD=0.62; teacher: M=0.73, SD=0.61) when compared with the UK comparison group (parent: M=0.43, SD=0.48; teacher: M=0.38, SD=0.58: parent: p<0.05; teacher: p<0.01). There was no appreciable difference between the UK comparison group and the Romanian non-ir group (parent: p=0.996; teacher: p=0.87). Effect of duration of deprivation Sizeable and significant differences in both the levels of I/O and the percentage in the abnormal range were found between those children who had experienced at least 6 months_ institutional care and those who had either experienced less than 6 months_ or no institutional care in Romania, or were adopted from within the UK. Overall, analysis of variance showed a highly significant difference in mean levels of I/O at age 11 across adoptee groups, as reported by both parents and teachers. A highly significant association was also found between adoptee group status and rates of I/O abnormality according to parent reports using a chi square test (see Table 1). Post hoc Tukey s tests found no difference in the level of I/O between the two late placed IR groups (6 to <24 and 24 months) or among the IR <6 month, the non-ir and the UK adoptee groups. The IR group aged 6 to <24 months was rated by parents and teachers as having significantly higher I/O scores than the IR <6 month, non-ir and the within-uk subsamples (parent: p<0.05; p<0.05; p<0.01; teacher: p<0.01; p<0.05; p<0.001 for the three groups, respectively). I/O in the IR 24 month group was significantly higher than the within-uk

7 J Abnorm Child Psychol (2008) 36: Table 1 I/O at age 11: mean scores, standard deviations and percentages in abnormal range I/O mean scores (SD) I/O % in abnormal range (n) Adoptee groups Parent Teacher Parent Teacher UK comparison Both sexes 0.43 (0.48) 0.38 (0.58) 6% (3) 8% (4) n=48 (parent) Male 0.51 (0.51) 0.52 (0.66) 6% (2) 12% (4) n=50 (teacher) Female 0.28 (0.40) 0.13 (0.22) 6% (1) 0% (0) Romanian non IR Both sexes 0.42 (0.49) 0.46 (0.52) 5% (1) 11% (2) n=20 (parent) Male 0.41 (0.37) 0.41 (0.52) 0% (0) 9% (1) n=19 (teacher) Female 0.43 (0.63) 0.54 (0.55) 11% (1) 13% (1) Romanian IR <6 months Both sexes 0.49 (0.49) 0.44 (0.47) 5% (2) 6% (2) n=42 (parent) Male 0.55 (0.53) 0.54 (0.48) 9% (2) 6% (1) n=35 (teacher) Female 0.42 (0.44) 0.35 (0.47) 0% (0) 6% (1) Romanian IR 6 <24 months Both sexes 0.84 (0.64) 0.92 (0.65) 25% (12) 25% (11) n=49 (parent) Male 0.89 (0.69) 1.08 (0.68) 32% (7) 33% (6) n=44 (teacher) Female 0.80 (0.60) 0.81 (0.61) 19% (5) 19% (5) Romanian IR months Both sexes 0.75 (0.67) 0.77 (0.59) 23% (9) 13% (5) n=40 (parent) Male 0.88 (0.77) 0.80 (0.60) 36% (5) 13% (2) n=40 (teacher) Female 0.69 (0.62) 0.75 (0.60) 15% (4) 12% (3) ANOVA (means); chi square (percentages) *p<0.05. **p<0.01. ***p<0.001 Both sexes F(4,194)=4.83** F(4,183)=6.98*** χ 2 (4)=13.93** χ 2 (4)=8.58; p=0.07 Male F(4,96)=2.50* F(4,89)=3.39* χ 2 (4)=13.39** χ 2 (4)=6.44 Female F(4,93)=3.03* F(4,89)=5.95*** χ 2 (4)=4.82 χ 2 (4)=4.61 sample according to teacher rating (p<0.05) but not parent ( p=0.07). When the non-ir group was excluded from the analysis the difference became significant ( p<0.05). When the two later placed adoptee groups were pooled and compared with the Rom IR <6 months group a t test showed there was a statistically significant difference between the groups by both reporters (parent: t(129)= 2.77; p<0.01; teacher: t(117)= 3.46; p<0.01). The two later placed IR adoptee groups were then combined and compared with a pooled subsample consisting of the three low risk groups: non-ir, within-uk and IR <6 months. T tests showed that the level of I/O in the combined >6 months IR group was significantly higher than that in the low risk subsample (parent: t(197)= 4.33; p<0.001; teacher: t(186)= 5.13; p<0.001).this pattern of results suggested a stepwise pattern of association between duration of deprivation and I/O at 11 years of age. The findings strongly suggested that the two later placed IR groups are at a significantly increased risk for elevated levels of I/O behavior in early adolescence but that no additional risk is added as one moves to the group with more than 24 months deprivation. Therefore, the subsequent analysis of continuity and phenotypic distinctiveness of I/O are focused on a merged risk sample of the IR 6 to <24 and 24 groups. Moreover, because the level of I/O in the non-ir group was similar to that reported in the within-uk group and the Romanian IR <6 month groups, the non-ir subsample, whose age at adoption is spread over the total range, has been excluded from the following analyses. Is There Individual Continuity in I/O Behavior? Overall mean scores remained fairly constant in the IR risk sample (aged 6 42 months at adoption) across time-points according to both parent and teacher reports (parent: age 6 M=0.88, age 11 M=0.80; teacher: age 6 M=1.00, age 11 M=0.85). In order to get an overall picture of continuity, or persistence, in I/O behavior in the later placed Romanian adoptee group, correlations across the two time-points were performed. There were highly significant correlations, according to both informants, between I/O at ages 6 and 11 years (parent report: r=0.67; p<0.001; teacher report: r=0.43; p<0.001). This was explored further using a categorical approach (see Materials and Methods). Figure 1 illustrates the moderate to strong individual continuity in I/O behavior between ages 6 and 11 years for reports of normality and impairment. Four-fifths of the children in the normal range for I/O at age 6 remained below cut-off at age 11 according to both parents and teachers. There was also moderate continuity in persistence of impairment with over half the children above cut-off according to parent and/or teacher at age 6 persisting to age 11. A similar pattern of results was found when parent and teacher reports were considered separately, although the drop off from 6 11 was largely owing to lower

8 392 J Abnorm Child Psychol (2008) 36: At 6 years: Above abnormal cut-off n =27 Below abnormal cut-off n = 42 n=14 (52%) n=13 (48%) n=9 (21%) n=33 (79%) At 11 years: Above abnormal cut-off n = 23 Below abnormal cut-off n = 46 Fig. 1 Continuity and change in I/O for Romanian institution-reared children aged 6 months or over at adoption I/O ratings from teachers at age 11, especially for girls. Eleven out of the 13 offset cases were female. Is Deprivation-related I/O Phenotypically Similar to ADHD/ Hyperactivity in Terms of Its Association with Gender, Conduct Problems, IQ and Executive Function? Table 2 sets out the associations between I/O in the IR risk subsample and conduct problems, intellectual and executive functioning, disinhibited attachment and gender. Gender Rates of both male and female persistent, early onset I/O were elevated when compared with population figures (see Fig. 2). We can also see that sex differences emerged in the Romanian sample in early adolescence. The sex ratio was 1:1.6, with boys having higher symptom levels. The picture was very similar when parent and teacher reports at age 11 were analyzed separately (ratio for parent report: 1:2; teacher report: 1:1.5). I/O and comorbidity with conduct problems Analysis of variance tests found there was a highly significant difference in the conduct scores across children in the normal, situational and pervasive I/O groups according to both parent and teacher reports (parent: F(2,76) =20.12; p<0.001; teacher: F(2,76) =36.59; p<0.001). Figures 3a and 3b show highly significant correlations ( p<0.001) between concurrent I/O and conduct problems, suggesting a strong contemporaneous association (parent report r=0.33; r=0.63; teacher report r=0.64; r=0.68, for ages 6 and 11, respectively). Developmental pathways from I/O and conduct problems Multiple regression was used to examine the independent contribution of conduct problems and I/O at 6 years to each domain at 11 years. According to parent reports, both I/O (β=0.23; p<0.01) and conduct problems (β=0.58; p< 0.001) at age 6 made an independent contribution to conduct variation at age 11. Conduct problems at age 6 also contributed to I/O variation at age 11 with a significant, but weak, association (β=0.15; p<0.05). Teacher data demonstrated no such associations. I/O and IQ The IR 6 42 month group as a whole had depressed IQ scores, with even the normal I/O subgroup scoring on average nearly15 IQ points (1 standard deviation in population norms) below the population mean of 100. There was no significant difference in IQ among the three I/O groups using analysis of variance testing (F(2,69)=0.52; p=0.59), and there was no correlation between I/O mean scores and WISC scores at age 11 (parent report: r= 0.12; p=0.33; teacher report: r= 0.15; p=0.20). I/O and executive dysfunction Interference control (Stroop test) and working memory performance (backwards digit Table 2 Pattern of associations at age 11 between I/O and conduct problems, IQ, executive function, disinhibited attachment and gender in the Romanian sample aged 6 months or over at adoption Phenotypic features: means (SD) Conduct problems IQ Executive function Disinhibited attachment Gender I/O impairment groups Parent report Teacher report WISC score Stroop task Digit span backwards Parent report Male Female Normal range 0.32 (0.27) 0.22 (0.25) (14.61) (8.5) 4.82 (1.80) 0.94 (1.24) 58% 69% n=51 n=51 n=46 n=44 n=45 n=51 n=18 n=33 Situational 0.73 (0.46) 0.65 (0.48) (13.67) (12.28) 3.74 (1.59) 1.90 (2.07) 26% 27% n=21 n=21 n=20 n=17 n=19 n=21 n=8 n=13 Pervasive 1.01 (0.36) 1.2 (0.14) (17.94) (11.21) 4.00 (2.00) 3.71 (1.60) 16% 4% n=7 n=7 n=6 n=5 n=6 n=7 n=5 n=2 ANOVA *p=0.05. **p<0.001 F(2,76)= 20.12** F(2,76)= 36.59** F(2,69)=0.52 p=0.59 F(2,63)= 3.05* F(2,67)=2.75 p=0.07 F(2,76)=11.40**

9 J Abnorm Child Psychol (2008) 36: % Romanian sample British norms (parent report) British norms (teacher report) Male Female Fig. 2 Percentages in abnormal range for I/O presented by gender in the Romanian-IR sample age 6 months or over at adoption span) were significantly correlated with I/O according to both parent and teacher reports: backwards digit span (parent report: r= 0.30; p<0.01; teacher report: r= 0.34; p<0.01); a Parent report Stroop (parent report: r=0.28; p<0.05; teacher report; r= 0.47; p<0.001). Using ANOVA, a significant difference was found among the three I/O severity groups (F(2,63)=3.05; p<0.05) on Stroop task performance, with over a 10 point difference in errors between those in the normal range for I/O according to parents and teachers and those with pervasive I/O. The difference across the three I/O groups approached, but did not reach, significance for the digit span task (F(2,67)=2.75; p=0.07), and was in the expected direction. Disinhibited attachment An analysis of variance test found a highly significant difference in levels of disinhibited attachment across the I/O cut-off groups (F(2,76)=11.40; p<0.001). Those with the highest I/O symptomatology were also rated by their parents as having significantly higher disinhibited behavior. Furthermore, there were highly significant bivariate correlations between disinhibited attachment and both parent and teacher reports of I/O (parent: r=0.46; p<0.001; teacher: r=0.25; p<0.05). A partial correlation between I/O and b Teacher report Age 6 Age 11 R 2 =0.54: F(2,124)=73.42;p<0.001 Age 6 Age 11 R 2 =0.20 F(2,104)=12.86;p<0.001 Inattention/ overactivity β=0.67; p<0.001 β =0.23;p<0.01 Inattention/ overactivity Inattention/ overactivity β =0.40; p<0.001 β =0.10;p=0.39 Inattention/ overactivity r=0.33; p<0.001 β =0.15;p<0.05 r=0.63; p<0.001 r=0.64; p<0.001 β =0.07;p=0.50 r=0.68; p<0.001 Conduct problems β =0.58; p<0.001 Conduct problems Conduct problems β =0.25; p<0.05 Conduct problems R 2 =0.47; F(2,124)=55.97; p<0.001 R 2 =0.10; F(2,104)=6.06; p<0.01 Prediction of conduct problems Prediction of I/O Concurrent I/O and CP correlation Fig. 3 Regression and correlation model of I/O and conduct problems in Romanian-IR sample

10 394 J Abnorm Child Psychol (2008) 36: disinhibited attachment was also highly significant (r=0.47; p<0.001) after controlling for the shared association with duration of deprivation. This raises the question of whether I/O, conduct problems and disinhibited attachment within this sample are overlapping but distinct constructs or different elements of the same underlying construct. To investigate this we carried out an exploratory factor analysis using the I/O, conduct problem and disinhibited attachment assessment items. The outcome measures seemed to distinguish the three domains as separate dissociable factors. The I/O items loaded together, along with one of the disinhibited attachment items: wandering off. This factor accounted for 43% of the variance. The other two disinhibited attachment items: too friendly with strangers/lack of differentiation and physical contact/lack of understanding social boundaries/personal space loaded together and accounted for a further 8% of the variance. The conduct items loaded on two additional factors, largely divergent according to whether the items tapped aggressive (e.g. bullies other children) or non aggressive (e.g. often tells lies) aspects of conduct disturbance and accounted for 12% and 9% of the variance, respectively. Discussion The current results help to identify a number of important characteristics of I/O as an early adolescent outcome of severe institutional deprivation. First, deprivation-related I/O persisted into early adolescence. However, high levels of I/O at 6 years only moderately predicted similarly high levels at 11 years. In general the findings support developmental continuity, in this as in other outcomes (Kreppner et al. 2007), evident in large differences between adoptee groups and high correlation in impairment from age 6 to 11. This persistence, despite the radical change in social environment following adoption, makes it highly unlikely that the effects are the result of a behavioral reaction to the poor conditions of the early environment, the influence of which one would expect to decrease with duration of time spent in good environments. Rather this is perhaps suggestive of some form of intraorganismic or fundamental neurobiological alteration. Rutter and O Connor (2004) hypothesized that persistent problems, such as I/O, following exposure to early severe adverse events, were the result of experience-adaptive biological programming, whereby the brain adapts to certain experiences during a critical period to optimize the specific conditions of that environment. This lends itself to the proposition that an alternative neuro-developmental pathway is initiated during an early critical period that is adapted to the stressful rearing environment (Teicher et al. 2003), a model that may hold some relevance for the persistent adverse effects presented above. Animal models support the existence of long lasting effects of early stress on brain development and on later psychological and behavioral functioning, including altered structure and function (e.g. HPA axis and associated brain structures) and neuro-chemical processes as it affects the processes of neurogenesis, synaptic overproduction and pruning and myelination (Teicher et al. 2003; McEwan 1999). One such model suggests that antenatal exposure to glucocorticoids (due to maternal stress or administration of a synthetic analogue during pregnancy) has long term effects on the HPA axis development and functioning of offspring and impacts on later locomotor activity in animals and ADHD-type behaviours in humans (Kapoor et al. 2007). Recent MRI work on a subsample of ERA participants is consistent with this model. Future research is needed to focus on the role of stress reactivity following early deprivation in developmental outcomes such as I/O. Second, and in contrast to the findings at age 6, the dose response relationship between I/O and duration of deprivation was marked by a clear step-like increase in risk at around 6 months of institutional deprivation consistent with a threshold model of early-deprivation-related risk. This is again consistent with accounts in which early adverse events need to occur within a critical developmental window for negative outcomes to follow (Bruer 2001). Due to the inevitable confound between age and duration of deprivation in the ERA study these models cannot be tested definitively using the current data. However, this finding does help to disentangle the notion of whether institutional rearing may in fact be a marker for some underlying genetic predisposition for problem behaviors such as I/O and also explore the potential confounding effect of prenatal risk factors on the prevalence of I/O in this sample. It is possible that parental ADHD or prenatal risk factors, such as low birth weight, maternal smoking or alcohol use during pregnancy or premature birth, known to be associated with ADHD in the general population, may have had some impact on the elevated levels and rates of I/O found in our sample. If selection into institutions reflected such a predisposition or if prenatal risk factors were driving the association between institutional deprivation and I/O then the increased risk for I/O should be spread across the adoptee age groups, and not just for those who experienced over 6 months_ deprivation. Moreover, if in fact it was the dose of deprivation that was the marker for genetic or prenatal risk then it follows that those children who experienced an extended period of deprivation would have greater genetic liability or prenatal adversity than earlier adopted children. One reason why it is unlikely that those children who were adopted at an older age would be differentially affected by such risk factors is that the ERA children could not be adopted until the fall of the Ceaus$escu regime, thereby largely avoiding the possibility that the children who

11 J Abnorm Child Psychol (2008) 36: resided longer in the institutions comprised those who had not been adopted sooner, possibly due to developmental or behavioral problems (which could be influenced by genetic makeup or prenatal adversity). However, such processes cannot be definitively tested and the potential influence of genetic predisposition and prenatal risk are important limitations to the current study. There are also several other potentially confounding factors that warrant mention here but are unfortunately outside the scope of the current paper and have been addressed in other papers by the ERA study. Factors such as differences in quality of care between individual children and between institutions (Castle et al. 1999), physical health status (Beckett et al. 2003) and post institution rearing environment may all potentially have had some impact on persistence and prevalence of I/O impairment. However, it is worth noting that the quality of care in the institutions ranged from poor to abysmal and that the post adoption rearing environments have not been found to mediate the impact of institutional deprivation on other areas of impairment, although this may be due to a lack of variation in range (see Colvert et al. in press; Kreppner et al. 2007). Our third main finding is that by 11 years deprivationrelated I/O impairment was more common in boys than girls, as is the case in nondeprivation-related I/O, with a similar gender discrepancy in prevalence rates to that seen in clinical and epidemiological populations. This was a different picture from that at age 6 where roughly equal numbers of boys and girls exhibited I/O difficulties. This shift in the sex ratio may reflect a developmental process whereby more general risk factors for I/O other than those specifically related to deprivation come into play as one moves further away in time from the institutional exposure. Fourth, deprivation-related I/O was associated with conduct problems. However, we need to be cautious about drawing a direct comparison between deprivation and nondeprivation-related I/O in this regard, because in nondeprived samples most evidence supports the model that I/O is a developmental precursor for conduct problems rather than the other way around (Burke et al. 2005). The current analysis suggest a more complicated reciprocal pattern with parent data showing both I/O leading to conduct problems and conduct problems leading to I/O. Additionally, teacher data does not support a developmental pathway from early I/O to later conduct problems at all. This lack of effect in teacher reports may be due to the fact that different raters are reporting at different ages. Individual teachers may have different tolerances to what they see as problematic behaviour in the classroom, especially at the different developmental stages. Parent reports are far more consistent over time. Furthermore, these mixed findings may be due in part to the behaviors being rated on questionnaire scales that are very good at picking up behavioral problems in general but are perhaps less proficient at picking up the development of multiple domains, particularly highly correlated ones. Fifth, I/O was associated with deficient executive functioning, at least as measured by the Stroop test of interference control, despite no association with IQ. This association was also reflected in the correlation between I/O and both our measures of executive functioning. However, the use of a risk subsample with depressed IQ scores overall suggests that the lack of effect of I/O status could be largely due to the overriding influence of duration of deprivation on IQ scores. Although one must be cautious about over-interpreting the finding of impaired executive function in relation to I/O, as it is based on only two tests, it does provide the first evidence that I/O in institutionally deprived samples bears the hallmark executive dysfunction found in ADHD and that it may also share elements of its psychopathophysiology (Willcutt et al. 2005). Studies with larger batteries of measures that investigate the functional neuro-anatomy of executive dysfunction in deprivation-related I/O are warranted. One key question is whether functional alterations in the fronto striatal circuits implicated in executive dysfunction in ADHD are also the basis for impairment in deprived samples, or are alterations in brain circuits more typically regarded as stress targets (hippocampus, amygdala etc) the main locus for impairment. Sixth, despite being dissociable constructs there was an overlap between I/O and disinhibited attachment, as has been identified in other institution reared samples (Roy et al. 2004). This overlap was not accounted for by the shared association with duration of deprivation. On the face of it this seems like a distinctive feature of deprivation-related I/O. At present, however, there is insufficient evidence from non-deprived samples to assess whether disinhibited attachment of the sort displayed by the deprived children in the current sample might also be present as an important clinical feature in at least a subsample of ADHD cases. In summary, the evidence suggests that I/O is a fairly stable domain of impairment for this group of children and the risk for I/O continues to be associated with institutional deprivation into early adolescence. This highlights the long lasting effects of the children s early adverse experience and institutional rearing. Furthermore, our analyses suggest that deprivation-related I/O shares a number of the features of I/O in non-deprived samples despite its different etiology and putative developmental mechanisms. This begs the question as to whether institutional deprivation should be seen as one (uncommon) route into a common disorder (ADHD) or whether deprivation-related I/O should be seen as a qualitatively different clinical phenotype with a distinct pathophysiology. One strategy for addressing this question scientifically involves hypothesizing a plausible, neuro-biological mechanism by which early institutional