Effect of Early Institutionalization and Foster Care on Long-term White Matter Development: A Randomized Clinical Trial

Importance  Severe neglect in early life is associated with compromises in brain development and associated behavioral functioning. Although early intervention has been shown to support more normative trajectories of brain development, specific improvements in the white matter pathways that underlie emotional and cognitive development are unknown.

Objective  To examine associations among neglect in early life, early intervention, and the microstructural integrity of white matter pathways in middle childhood.

Design, Setting, and Participants  The Bucharest Early Intervention Project is a randomized clinical trial of high-quality foster care as an intervention for institutionally reared children in Bucharest, Romania, from 2000 through the present. During infancy, children were randomly selected to remain in an institution or to be placed in foster care. Those who remained in institutions experienced neglect, including social, emotional, linguistic, and cognitive impoverishment. Developmental trajectories of these children were compared with a group of sociodemographically matched children reared in biological families at baseline and several points throughout development. At approximately 8 years of age, 69 of the original 136 children underwent structural magnetic resonance imaging scans.

Main Outcomes and Measures  Four estimates of white matter integrity (fractional anisotropy [FA] and mean [MD], radial [RD], and axial [AD] diffusivity) for 48 white matter tracts throughout the brain were obtained through diffusion tensor imaging.

Results  Significant associations emerged between neglect in early life and microstructural integrity of the body of the corpus callosum (FA, β = 0.01 [P = .01]; RD, β = −0.02 [P = .005]; MD, β = −0.01 [P = .02]) and tracts involved in limbic circuitry (fornix crus [AD, β = 0.02 (P = .046)] and cingulum [RD, β = −0.01 (P = .02); MD, β = −0.01 (P = .049)]), frontostriatal circuitry (anterior [AD, β = −0.01 (P = .02)] and superior [AD, β = −0.02 (P = .02); MD, β = −0.01 (P = .03)] corona radiata and external capsule [right FA, β = 0.01 (P = .03); left FA, β = 0.01 (P = .03); RD, β = −0.01 (P = .01); MD, β = −0.01 (P = .03)]), and sensory processing (medial lemniscus [AD, β = −0.02 (P = .045); MD, β = −0.01 (P = .04)] and retrolenticular internal capsule [FA, β = −0.01 (P = .002); RD, β = 0.01 (P = .003); MD, β = 0.01 (P = .04)]). Follow-up analyses revealed that early intervention promoted more normative white matter development among previously neglected children who entered foster care.

Conclusions and Relevance  Results suggest that removal from conditions of neglect in early life and entry into a high-quality family environment can support more normative trajectories of white matter growth. Our findings have implications for public health and policy efforts designed to promote normative brain development among vulnerable children.

Trial Registration  clinicaltrials.gov Identifier: NCT00747396

Many aspects of postnatal brain development depend heavily on experience. Consequently, serious violations of what is termed the expectable environment (ie, experiences that all members of our species should expect to encounter)¹ can lead to profound changes in neural development. Institutional rearing represents a profound violation of the expectable environment in that children typically experience high ratios of children to caregivers, limited access to language and cognitive stimulation, and insufficient caregiving. Not surprisingly, institutionally reared children often show compromises in brain development and associated behavioral functioning.^2-4

Recent investigations using diffusion tensor imaging (DTI) have demonstrated significant associations between institutional neglect and microstructural alterations in white matter. Alterations are widespread and have included limbic and paralimbic pathways,^5-7 frontostriatal circuitry,^7-9 and sensory-processing pathways.⁷ Although findings are compelling, these studies share a methodological weakness associated with the potential for sample biases: institutionalized children selected for international adoption may differ developmentally from those not selected. One potential example is that the IQs of internationally adopted children often fall within the normal range, whereas the IQs of comparably aged children who remain in institutions often fall 2 to 3 SDs below the mean.^6,10,11

Randomized clinical trials involving early intervention can overcome this methodological issue and uncover associations not confounded by selection biases. Improved total white matter content during middle childhood has recently been demonstrated in children randomly assigned to enter into a responsive family setting relative to those who remained in institutions.¹² However, the microstructural changes that underlie these global white matter improvements have not yet been elucidated.

The present study investigated the white matter integrity of 3 groups of children who participated in the Bucharest Early Intervention Project (BEIP), a randomized clinical trial of Romanian infants reared in institutional settings. During infancy, children were randomly assigned to remain in the institution or to be removed from the institution and placed in high-quality foster care. Developmental trajectories were compared with those of a group of demographically matched children reared in biological families. We hypothesized that institutionally reared children would show abnormalities in white matter integrity throughout the brain, specifically in regions supporting cognitive and emotional regulation. We expected that white matter compromises would be most severe for children who remained in the institution. We also hypothesized that institutionally reared children placed in foster care would show evidence of remediation in specific fiber tracts as a result of early intervention.

The institutional review boards from the University of Maryland, College Park; Boston Children’s Hospital, Boston, Massachusetts; and Tulane University, New Orleans, Louisiana, approved all study procedures, as did an institutional review board established in Romania. In addition, written informed consent was obtained from each of the 6 local Commissions for Child Protection in Bucharest and/or the biological parents when possible. The full study protocol can be found in the trial protocol in Supplement 1.

The BEIP is, to our knowledge, the first-ever randomized clinical trial of foster care as an intervention for early institutionalization. At approximately 2 years of age, 136 children who had spent more than half of their lives in institutions in Bucharest, Romania, were recruited for the study (Figure). The BEIP core group (involving principal investigators and original staff members of the study) performed randomization procedures. Each child was assigned a number from 1 to 136, which was written on a piece of paper and placed into a hat. Slips of paper were then drawn from the hat one at a time. The first number selected was assigned to the care as usual group, and the next number selected was assigned to the foster care group. This process was repeated until all 136 children were assigned to either group. Enrollment and randomization began in 2001. After the baseline assessment, follow-up assessments took place when children reached 30 months, 42 months, 54 months, 8 years, and 12 years of age. A third group of age- and sex-matched children reared in their biological families in Bucharest (never institutionalized) was used as a comparison group.¹³ (See the eTable in Supplement 2 for characteristics of the participants at baseline.)

The BEIP drew from all 6 institutions for abandoned children in Bucharest, Romania. Although there was some variability within and between institutions in terms of the quality of the care environment, all where characterized by a high ratio of children to caregivers (on the order of 12 children for every caregiver). In addition, all were characterized by impoverished social, emotional, linguistic, and cognitive experience. Finally, there were limited opportunities for infants to practice their motor skills, as most infants were confined to cribs during their first 1 to 2 years of life.

At the onset of the study, foster care was almost nonexistent in Bucharest, Romania. Historically, institutional care centers were used as the standard form of care for abandoned children. The BEIP core group developed a network of foster families (n = 56) who cared for the 68 children randomized into foster care. Manuals for foster care recruitment and training were developed by and for citizens of Romania. Three social workers provided assistance to BEIP foster caregivers. Social workers received initial training and consulted with experienced mental health practitioners in the United States on a weekly basis and had frequent and regular contact with foster parents. Specifically, social workers monitored the relationship between the foster caregivers and children, promoted high-quality parent/child attachment relationships, provided guidance for behavioral management when necessary, and informed foster parents on the highly specialized needs of the children in their care. Overall, social workers encouraged foster caregivers to develop committed and loving relationships with the children placed in their care. Many foster caregivers eventually adopted the children placed in their care despite their having to give up a salary and assume all costs related to caring for their child. Foster placements were highly stable from randomization. Children were placed into foster care between 5 and 31 months of age. At an assessment that took placed when children reached 54 months of age, 87% of children remained with their original foster family. At this point in the intervention, the local government in Bucharest, Romania, assumed financial and administrative support of foster parents and children.

Diffusion tensor imaging data from 69 participants (aged 8-11 years) were selected for the tract-based spatial statistics analysis to investigate potential white matter abnormalities due to institutional rearing during early development. Participants included 23 children randomized from the institution into foster care (mean [SD] age, 9.87 [0.63] years), 26 children randomly assigned to remain in institutional care (mean [SD] age, 9.69 [0.93] years), and 20 children who had never been in institutional care (mean [SD] age, 9.80 [0.52] years). We found no statistically significant differences in children’s ages (P = .69) or sex (P = .35) across groups at the time of the magnetic resonance imaging (MRI) assessment.

Diffusion tensor imaging was performed on a 1.5-T scanner (Siemens and General Electric) using a single-shot echoplanar imaging sequence with twice-refocused spin echoes. The scanning variables for DTI acquisitions were as follows: repetition/echo times, 8600/100 milliseconds; section thickness, 2.3 mm with no gap and a total of 55 sections for whole-brain coverage; data matrix, 208 × 208; and field of view, 240 ×240 mm. Diffusion-weighted images were acquired along 30 noncollinear and noncoplanar directions with a b value of 1000 s/mm² and 2 images with a b value of 0 s/mm².

Tensor and tensor-derived parametric maps for fractional anisotropy (FA) and mean (MD), radial (RD), and axial (AD) diffusivity were first estimated using an analysis tool for brain imaging (DTIFIT tool in the FSL package; FMRIB Analysis Group). Maps were then fed into the tract-based spatial statistical tool to generate a white matter skeleton.¹⁴ Considering the ages of participants in the BEIP, a study-specific template in the standard space, instead of a standard adult brain template, was created in a 2-step approach¹⁵ for the tract-based spatial statistical analysis used in this study.

The DTI atlas¹⁶ from the Laboratory of Brain Anatomical MRI at The Johns Hopkins University included in the FSL package (the ICBM-DTI-81 White Matter Atlas; hereinafter referred as the WM Atlas) was chosen as a template to facilitate identification of major white matter structures. Forty-eight tracts from the WM Atlas were identified for analyses in the present study (Table 1 provides a complete list of tracts) using nomenclature and names established by Mori and van Zijl.¹⁷ Mean FA, MD, RD, and AD values across all voxels for each of the 48 tracts as defined by the WM Atlas were calculated. An individual mean DTI index for each tract was extracted per participant using the FSL package.

All statistical analyses were conducted using the software R (http://www.r-project.org). We compared the DTI data between groups primarily using linear regression models. The analysis examined group differences with children categorized as falling in their originally assigned care as usual or foster care groups. However, over time, some children originally assigned to these groups underwent changes in living arrangements.¹³ Therefore, the analysis provides a conservative estimate of the effect of the early intervention on white matter microstructure.

We first developed linear regression models to investigate correlations between structural alterations in white matter (the outcome) and histories of institutional rearing or participant group (the independent variable; categorized as 1 for the care as usual group, 2 for the foster care group, and 3 for the never-institutionalized group). Individual models were developed for each tract and each DTI variable. The relatively small samples limited the development of larger models. Because this analysis aimed to assess the sensitivity of individual tracts and DTI variables, the issue of multiple comparisons was not of concern, and associations were considered significant at P < .05. Multinomial regression models were also developed to compare pairs of tracts across groups. These models used the never-institutionalized group as the reference group and compared the care as usual and foster care groups with it. The significance level was adjusted for these 2 comparisons in the models.

We identified 4 tracts in which FA was statistically distinct in the 3 groups: the body of the corpus callosum, left external capsule, right external capsule, and right retrolenticular internal capsule. We identified 4 tracts in which RD was statistically distinct in the 3 groups: the body of the corpus callosum, right cingulum, left external capsule, and right retrolenticular internal capsule. We identified 4 tracts in which the AD was statistically distinct in the 3 groups: the right anterior corona radiata, right fornix crura, right medial lemniscus, and left superior corona radiata. Finally, we identified 6 tracts in which the MD was statistically distinct in the 3 groups: the body of the corpus callosum, right cingulum, left external capsule, right medial lemniscus, right retrolenticular internal capsule, and left superior corona radiata (Table 2). Significant associations emerged between neglect in early life and microstructural integrity of the body of the corpus callosum (FA, β = 0.01 [P = .01]; RD, β = −0.02 [P = .005]; MD, β = −0.01 [P = .02]) and tracts involved in limbic circuitry (fornix crus [AD, β = 0.02 (P = .046)] and cingulum [RD, β = −0.01 (P = .02); MD, β = −0.01 (P = .049)]), frontostriatal circuitry (anterior [AD, β = −0.01 (P = .02)] and superior [AD, β = −0.02 (P = .02); MD, β = −0.01 (P = .03)] corona radiata and external capsule [right FA, β = 0.01 (P = .03); left FA, β = 0.01 (P = .03); RD, β = −0.01 (P = .01); MD, β = −0.01 (P = .03)]), and sensory processing (medial lemniscus [AD, β = −0.02 (P = .045); MD, β = −0.01 (P = .04)] and retrolenticular internal capsule [FA, β = −0.01 (P = .002); RD, β = 0.01 (P = .003); MD, β = 0.01 (P = .04)]).

Next, separate linear regression models examined whether associations among each of these 18 DTI values (4 FA, 4 RD, 4 AD, and 6 MD values) continued to be associated with group membership (foster care, care as usual, or never institutionalized) when controlling for covariates (age, birth weight, and intracranial volume). Identified tracks and corresponding DTI variables continued to be statistically distinct in the 3 groups even when the covariates were included in the model. Overall, these covariates were not significantly associated with the DTI variables except for birth weight, which was positively associated with FA for the body of the corpus callosum. However, the positive association between group membership and FA of the body of the corpus callosum remained significant even when controlling for birth weight.

Next, we tested whether combinations of tracts were more strongly associated with group membership when compared with each tract as an independent predictor. Pairs of uncorrelated tracts were tested in multinomial logistic regression models with group as the outcome and DTI variables as predictors (Table 3 shows pairs of tracts that were not significantly correlated with each other for each DTI variable). Results of the multinomial logistic regressions revealed that no tract pairs were distinct in combination in the 3 groups. This lack of association could be a result of the small sample size; however, in the absence of a larger sample to verify the lack of combinatorial tract correlations with group membership, our results suggest that associations between each DTI variable for each tract and group occurred independently rather than in combination with other tracts.

Several multinomial regression models showed that, in some cases, values for certain tracts were statistically significantly associated with the logarithm of odds of belonging to the care as usual group relative to the never-institutionalized group but were not significantly associated with the logarithm of odds of belonging to the foster care group relative to the never-institutionalized group, suggesting an intervention effect. This evidence for remediation in the foster care group but not the care as usual group was observed for FA values in the left external capsule; FA values in the right external capsule; FA, MD, and RD values in the retrolenticular internal capsule; MD and RD values in the right cingulum; AD values in the right anterior corona radiata; AD values in the left superior corona radiata; MD and AD values in the medial lemniscus; and (at a trend level) AD values in the right fornix crura (Table 4).

Finally, we examined whether variations in the timing of the intervention (ie, entry into foster care) predicted white matter integrity during middle childhood. Because the age of placement in foster care was associated with the age at the MRI scan (r = 0.89; P < .001), child’s age was entered as a covariate in the analyses. We found no significant associations between intervention timing and white matter integrity when accounting for the effects of child’s age at the time of the scan.

This investigation is the first, to our knowledge, to examine the effects of neglect in early life on the organization of white matter microstructure within the context of a randomized clinical trial of foster care as an intervention for early institutionalization. The randomized design is a critical strength of this investigation because it allows for the control of potential selection biases encountered in previous investigations involving internationally adopted youth. Results from this study extend prior knowledge by further delineating white matter tracts affected by extreme neglect in early life. From these data, we suggest that removal from conditions of severe neglect in early life and entry into a high-quality family environment may support more normative trajectories of white matter growth in the long term.

Our results revealed that neglect in early life was associated with alterations in white matter microstructure throughout the brain, specifically involving the body of the corpus callosum, cingulum, fornix, anterior and superior corona radiata, external capsule, retrolenticular internal capsule, and medial lemniscus. The foster care group did not significantly differ from the never-institutionalized group in measurements of these tracts, with the exception of the body of the corpus callosum and superior corona radiata. These findings suggest a potential for remediation of specific white matter pathways for children removed from institutional care and placed in responsive families early in life.

The BEIP intervention focused on facilitating high-quality parent/child attachment relationships between the institutionally reared children and their foster care providers. As part of the program, foster parents were encouraged to develop responsive, committed relationships with their child; were educated on the child’s specialized cognitive and emotional needs; and were provided guidance on behavioral management strategies to support the child’s optimal development. A previous study¹² demonstrated evidence of intervention-associated improvements in total white matter volume among institutionally reared children placed in foster care. The present results delineate the specific white matter tracts that may contribute to the global improvements in white matter changes. Prior work also has demonstrated that caregiving-based early intervention programs can support more normalized white matter development among children who are exposed to prenatal risk.^18,19 Our results suggest a similar potential for recovery in children exposed to extreme early adverse conditions in the postnatal period.

Evidence presented in this study introduces several questions for future research. First, assessments of white matter microstructure occurred approximately 6 years after children were randomized into responsive family settings. Therefore, the specific timing and rate of white matter improvements among children in foster care is unknown. White matter increases linearly across development, and experience-expectant and experience-dependent processes drive its growth and organization.²⁰ Improvements in white matter integrity could have occurred from appropriate, experience-expectant, caregiving input at sensitive periods of brain development in early childhood and/or from ongoing exposure to enriching, experience-dependent experiences throughout the course of development.

The specific neural changes that contribute to these quantitative estimates of microstructural improvements are also unknown. Early-life alterations in neural pruning and axonal organization may have contributed to these long-term white matter patterns. However, changes in the overall rates of the myelination that occurs across the course of development may also contribute to the group differences observed in this study. Future investigations involving longitudinal assessments of neural development will be critical for identifying the specific neural properties that subserve our observed long-term changes. Understanding these specific trajectories of white matter changes may have important public health implications regarding the timing, duration, and format of early interventions delivered to at-risk children.

In terms of the specific white matter tracts, children in the care as usual and foster care groups showed reduced integrity (decreased FA and increased RD and MD) in the body of the corpus callosum when compared with children reared in family settings. Alterations in this region are consistent with prior work demonstrating a smaller volume of the corpus callosum^21,22 and reduced microstructural integrity^22-24 among individuals exposed to maltreatment in family settings. The corpus callosum is the largest myelinated fiber tract in the brain and supports interhemispheric transmission of neural information. Abnormalities in the corpus callosum have been associated with psychiatric and developmental disorders, including attention-deficit/hyperactivity disorder²⁵ and cognitive and language delays.²⁶ Symptoms related to attention-deficit/hyperactivity disorder in children exposed to deprivation seem especially persistent, even in children assigned to the foster care intervention.²⁷ Long-term reductions in the integrity of the body of the corpus callosum for children in the care as usual and foster care groups may subserve these pervasive patterns of neurocognitive risk.

Two white matter tracts involved in limbic circuitry were significantly associated with institutional rearing in this study. The cingulum, a collection of white matter fibers that runs along the cingulate gyrus and projects to the entorhinal cortex, supports communication between the frontal and limbic regions of the brain.^28,29 The fornix crus, a flat band of efferent fibers in the posterior portion of the fornix, projects to the dorsal regions of the hippocampus. Reduced integrity in these regions, manifesting specifically as increased RD and MD for the cingulum and reduced AD in the body of the fornix, has been observed among individuals exposed to adverse early rearing conditions in several prior investigations.^6,7,30 The integrity of these regions also has been linked with increased difficulties in externalizing,^6,31 internalizing,^30,32 inattention,³³ and spatial planning.⁷ A remaining question is whether these white matter disruptions underpin similar difficulties observed previously in the institutionally reared children in the present sample.^27,34,35

Histories of institutionalization were also associated with compromised integrity of tracts involved in frontostriatal circuitry, specifically manifesting as decreased AD in the right anterior corona radiata, decreased AD and MD in the left superior corona radiata, decreased FA in the left capsule, and decreased FA and increased RD and MD in the right external capsule. The corona radiata is a bundle of white matter fibers that connect the cortex with the thalamus, basal ganglia, and spinal cord. The anterior portion connects the anterior cingulate cortex with the striatum, and disruptions to this portion of the corona radiata are consistent with the findings of a prior investigation involving institutionally reared children.⁷ Functionally, this tract has been implicated in cognitive, emotional, and behavioral regulation.^36,37 More specifically, poorer integrity in this tract has been associated with spatial planning difficulties among institutionally reared children.⁷ The external capsule is a series of white matter tracts that connect the cortex to the striatum. Although the specific function of the external capsule is largely unknown, reduced integrity has been associated with risk for addiction and substance abuse, compromised regulatory skills, and poor cognitive control.³⁸ Understanding the functional correlates of the reduced integrity of these tracts for children in the current sample will be an important direction for future work.

Finally, neglect in early life was also associated with alterations in 2 white matter tracts implicated in basic sensory processing. These tracts included the right retrolenticular internal capsule and the right medial lemniscus. The retrolenticular internal capsule contains fibers involved in the visual system. Unexpectedly, histories of institutional neglect were associated with higher FA and lower MD and RD. The medial lemniscus is a major afferent pathway that carries sensory information from the brainstem to the thalamus. Results revealed positive associations between neglect in early life and MD and AD values in this region. Reduced integrity in the medial lemniscus may result from insufficient sensory input experienced at critical points in neural development and may be associated with lower-level difficulties in sensory processing.

The inclusion of multiple DTI variables in the analytic approach is a strength of this study because the examination of MD, RD, and AD variables may yield a more comprehensive understanding of specific white matter properties.³⁹ We observed microstructural alterations of white matter tracts across all 4 variables, suggesting that neglect in early life may be associated with a variety of alterations in white matter development involving fiber density, membrane structure, myelination, axonal organization, and projection.

Results from this study contribute to growing evidence that severe neglect in early life affects the structural integrity of white matter throughout the brain and that early intervention may support long-term remediation in specific fiber tracts involved in limbic and frontostriatal circuitry and the sensory processes. Our findings have important implications for public health related to early prevention and intervention for children reared in conditions of severe neglect or adverse contexts more generally. Understanding links between these white matter profiles and neurocognitive or psychiatric functioning will be an important aim for future work and will shed light on mechanisms underlying risk and resiliency among children exposed to adverse early rearing conditions.

Accepted for Publication: November 10, 2014.

Corresponding Author: Charles A. Nelson, PhD, Boston Children’s Hospital, Harvard Medical School, Harvard Graduate School of Education, 1 Autumn St, Cambridge, MA 02115 (charles_nelson@harvard.edu).

Published Online: January 26, 2015. doi:10.1001/jamapediatrics.2014.3212.

Author Contributions: Dr Bick had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Fox.

Acquisition, analysis, or interpretation of data: Bick, Zhu, Stamoulis, Zeanah, Nelson.

Drafting of the manuscript: Bick, Stamoulis, Fox, Zeanah.

Critical revision of the manuscript for important intellectual content: Zhu, Stamoulis, Fox, Zeanah, Nelson.

Statistical analysis: Bick, Zhu, Stamoulis, Fox, Nelson.

Obtained funding: Fox, Zeanah.

Administrative, technical, or material support: Zhu, Fox, Zeanah, Nelson.

Study supervision: Zeanah.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the John D. and Catherine T. MacArthur Foundation; by the Binder Family Foundation; by the foundation Help the Children of Romania, Inc; and by grant MH091363 from the National Institute of Mental Health (Dr Nelson).

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.