Growth and relations between auxology, caregiving environment and cognition in socially deprived Romanian children randomized to foster vs. ongoing institutional care

Abstract

Objective

To determine the effects of improved nurture compared with institutional care on physical growth and the relations between growth and cognitive development.

Design

A randomized controlled trial beginning at baseline (mean 21.0, range 5.4–32 months) with follow-up at 30, 42 and 54 months of age.

Setting

Institutionalized and community children in Bucharest, Romania

Participants

136 institutionalized though otherwise healthy children from six orphanages and 72 typically developing, never-institutionalized children.

Intervention

Institutionalized children were randomly assigned to foster care or to receive institutional care as usual.

Outcome Measures

Auxology and measures of intelligence over time.

Results

Growth in institutionalized children was compromised, particularly in infants <2,500 grams. Mean height and weight, though not head size, increased to near normal within 12 months in foster care. Significant independent predictors for greater catch-up in height and weight included age <12 months at randomization and higher caregiving quality, particularly caregiver sensitivity and positive regard. At 54 months, birth weight < 2500 grams was associated with a 11.0±3.7(standard error) point lower IQ and each incremental increase of one in standardized height scores between baseline and 42 months was associated with a 6.2±3.0 point increase in IQ (p<.05).

Conclusions

Foster care had a significant effect on growth particularly with early placement and high quality care. Growth and IQ in low birth weight children are particularly vulnerable to social deprivation. Catch-up growth in height under more nurturing conditions is a useful correlate of caregiving quality and cognitive abilities.

Trial Registration

clinicaltrials.gov Identifier: NCT00747396

Introduction

A syndrome of poor growth in socially deprived children has been recognized since the eponymous Kasper Hauser was discovered stunted and developmentally delayed outside the gates of Nuremberg in 1828.¹ Several sub-types have been described, ^2–5 though all share two characteristics: otherwise unexplained growth failure occurring in association with socially stressful conditions and significant catch-up when a child’s caregiving environment improves.³ Understandably, studies to date have relied on convenience samples of children referred to specialists for evaluation of short stature, ^{2, 3, 5, 6} entering the child protection system ^7–9 or placed for adoption from institutional care settings.^{4, 10–12} Under these circumstances basic social and medical information is lacking and it has been virtually impossible to ascertain the type, severity or duration of adversity experienced by the child. This absence of randomized, controlled studies examining environment and growth may account for reported variations in the response of children to deprivation and social intervention, as well as the variety of growth failure subtypes described.

The Bucharest Early Intervention Project (BEIP), the first randomized controlled study of foster vs. institutional care, offered a unique opportunity to study growth within otherwise healthy institutionalized and never institutionalized Romanian children. These subjects had measures of growth, caregiving environment, behavior, brain and cognitive development assessed before and at intervals following randomization of those institutionalized to foster vs. institutional care as usual.^13–32 Contrasting these children with matched, never institutionalized controls, we explored biologic and environmental factors contributing to growth failure and recovery and investigated the relations between growth and cognitive outcome.

Methods

Study Design

Subject selection has been discussed in detail in past publications but will be summarized briefly.^{18, 22, 27} Enrollment, group assignments and follow-up of subjects through 54 months have been published previously (Fig. 1).¹⁸

Figure 1.

Group status at age 54 months for children living in Romanian institutions who were assigned to usual care or foster care.

Children in Institutions (IG)

All children < 32 months of age living in institutions for infants in all six sectors of Bucharest, other than those scheduled for adoption, were included (n=187). Children (n=51) with serious handicapping conditions were excluded e.g., genetic syndromes, facial features (3 or 4) indicative of high risk of prenatal alcohol exposure,^{33, 34} severe microcephaly, cerebral palsy or a high suspicion of bilateral hearing loss. Subjects were recruited from February through June 2001.

Care as Usual (CAU) and Foster Care Groups (FCG)

Following identification and baseline assessment of the institutionalized cohort (IG) (n=136), equal numbers were randomly assigned to remain within institutional care and receive care as usual (CAU), or to foster care (FCG). Eleven of the children originally cleared for participation in the study were later determined to have conditions that met exclusion criteria and were eliminated from analyses.²² Because government sponsored foster care was unavailable when our study commenced, we created our own foster care program.^{23, 27} Irrespective of changes in caregiving environment over the duration of the study, an intent-to-treat approach was followed, whereby all analyses we report are based on children’s original group assignments. Thus, our findings represent a conservative estimate of the response to intervention

Never-Institutionalized Group (NIG)

Children (n=72) were recruited from community pediatric clinics and were born at the same hospitals as the institutional children. They were living with their birth family, had no history of institutional care and were approximately matched on age and gender to CAU and FCG.

Ethical Considerations

The rationale behind the study, justification and critique of the randomized controlled design, and the approval process and procedures put in place to safeguard these children have been discussed in depth in previous publications.^{18, 27, 35–37} Approval was obtained from the Institutional Review Boards of the home institutions of the principal investigators (Charles Nelson, Charles Zeanah, and Nathan Fox).

Data Collection

Assessments of cognitive level and caregiving environment were obtained at three time points, baseline [21.0±7.4(s.d.), range 5–32 months], 30 and 42 months as previously described.^{18, 22} Birthweight was obtained from record review (IG) or from parent report (NIG) and was available for 90% of the children from IG and 94% of the children from NIG. Information on gestational age at birth for IG was either not available or deemed unreliable so it was impossible to determine whether low birth weight (LBW) infants (<2,500 grams) were appropriate or small for gestational age (SGA). In addition to baseline, 30 and 42 month auxology, measurements were scheduled monthly in CAU and FCG. Cognitive testing was also obtained at 54 months. All reported p-values are two sided. Data were analyzed using SPSS (Version 11.0 for Macintosh) and SAS (version 9.1) software.

Physical growth

Length/height, weight and weight-for-height were converted to age-standardized scores (z-scores) based on CDC 2000 data ³⁸ using the formula

$$z=(x-\mathrm{\mu })/\mathrm{\sigma },$$

where x is the raw measurement, μ is the age-specific mean, and σ is the age-specific standard deviation. Using z-score measures standardized sizes, thus making them comparable across age ranges. Occipital frontal circumference (OFC) z-scores were calculated using the standardized data (0–18 years of age) of Roche and colleagues ³⁹

Cognitive development and caregiving environment

Developmental quotients (DQ) at baseline, 30 and 42 months were based on the Bayley Scales of Infant Development II Mental Developmental Index. ^{22, 40} The Wechsler Preschool Primary Scale of Intelligence-II (WPPSI) was used at 54 months.⁴¹ The Observational Record of the Caregiving Environment (ORCE) ⁴² was adapted and used to assess a child’s caregiving experience in either the institution or family settings as previously described.²² The Caregiving Quality Score (CQS) was obtained by averaging five qualitative ORCE scales [i.e. detachment (reversed) flat affect (reversed), positive regard for child, sensitivity and stimulation of development], each of which received a rating from 1 (not at all characteristic) to 4 (highly characteristic). CQS ranged from 1 (lowest possible score) to 4.

Results

Control Analyses

Children in NIG were primarily ethnic Romanians (Table I). Birthweight was lower for IG children and the incidence of LBW (<2,500 grams) was higher than in NIG. At baseline, children randomized into CAU and FCG groups did not differ from one another in any of these parameters.

Table I.

Demographic, growth, caregiving quality and cognitive scores (± s.d.) for the institutionalized and community groups pre- and post-randomization.

Child Characteristics	Institutionalized (N=125)		Never Institutionalized (N = 72)
Age (months) (SD)	21.0 (7.4)		19.3 (7.1)
Ethnicity
Romanian	55%*		91%
Roma (Gypsy) Other/Unknown	45%		9%
Female	50%		57%
Birthweight (grams)	2,834 (596) *		3,333 (459)
< 2500 grams	24%*		3%
Percent of life Institutionalized	62.9% (25.1) *		0
Baseline height z-score	−.84 (.86) ‡		.13 (.91)
% < −2	9%*		2%
Baseline weight z-score	−1.23 (1.08) *		−.05 (1.00)
% < −2	25%*		0%
Baseline head circumference z-score	−1.10 (.99) *		−.15 (.86)
% < −2	17%†		2%
Baseline weight-for-height z-score	−.67 (1.14) *		.16 (.96)
% < −2	16%†		2%
Post Randomization Group	CAU (62)	FCG (63)
Baseline Caretaking Quality Score	2.06 || (.62)	2.29|| (.57)	2.82 (.53)
30 Month Caretaking Quality Score	2.46 §, ||,,¶ (.63)	2.75§ (.52)	2.86 (.55)
42 Month Caretaking Quality Score	2.53 § (.58)	2.79 § (.57)	2.75 (.59)
Mean 30/42 Month Caretaking Quality Score	2.49 §, ||,,¶ (.48)	2.76 § (.48)	2.81 (.49)
Cognitive Measures
Baseline DQ	72.5|| (13.2)	76.1|| (13.1)	103.8 (11.3)
DQ 42 months	77.1§, ||,,¶ (13.3)	85.7§, || (14.2)	103.4 (11.8)
WPPSI Full Scale IQ −54 Months	73.3 || ¶ (13.1)	81.0 §,|| (18.5)	109.3 § (21.2)

t- or χ² tests

^*p ≤ .001 vs. NIG

^†p < .01 vs. NIG

^‡p < .05 vs NIG

^§p < .05 vs baseline scores

One-way ANOVA post-hoc tests

^||p < .05 vs. NIG

^¶p < .05 vs. FCG

Auxology at Baseline

At baseline, all physical measurements were significantly smaller and z-scores < −2 more frequent in IG vs. NIG. The CQS was significantly lower in IG vs. NIG (Table I) and increased with age [y = 1.8 + .02 (age in months) F(1,114) = 6.00, p = .016, R² = .05]. The prevalence of wasting (weight-for-height z-scores < −2) in IG was high (16%) ⁴³ and was especially common in younger children [25% <12 mo vs. 6% ≥ 24 mo, χ²(1) = 7.14, p <.01].

Multiple regression models accounted for significant variance in baseline z -scores for height, weight, OFC and weight-for-height (Table II). Significant unique predictors of lower z-scores at baseline included lower birth weight (height, weight, OFC and weight-for-height), older age (height), non-Romanian ethnicity (OFC) and younger age (weight-for-height). Within IG, LBW children were significantly more growth impaired than children with birth weights ≥2.5 kg in all parameters other than OFC (p = .08) (Table III). Both birth weight groups were significantly different vs. NIG. Institutionalized children of Roma/other/unknown ethnicity had smaller OFC z-scores at baseline than children who were of Romanian ethnicity [−1.43±.83 (s.d.) vs −.83(1.03), t(118)=3.43, p=.001]. However, birth weight, baseline CQS and baseline DQ did not differ between these two ethnic groups making it more likely that this difference was an inherent auxologic characteristic rather than the result of pre- or post-natal factors.

Table II.

a) Multiple regression of baseline measures on gender, ethnicity, birth weight, age and caregiving quality in the institutional group (IG) at baseline.

	Baseline Height z-score	Baseline Weight z-score	Baseline OFC z-score	Baseline Wt/Ht z-score
	β (SE β)	β (SE β)	β (SE β)	β (SE β)
Intercept	−1.53 (.46) *	−3.64 (.62) *	−2.17 (.57) *	−3.04 (.70) *
Gender (f = 0, m = 1)	−.17 (.15)	−.04 (.20)	.04 (.18)	.08 (.22)
Ethnicity (Romanian = 0, other = 1)	−.07 (.15)	−.15 (.20)	−.50 (.18) †	−.09 (.23)
Birth Weight (kg.)	.67 (.13) *	.79 (.17) *	.36 (.16) ‡	.47 (.20) ‡
Caregiving Quality Score	−.24 (.13)	−.09 (.17)	.05 (.15)	.15 (.19)
Age (months)	−.03 (.01) ‡	.02 (.01)	.01 (.01)	.03 (.02)
Adjusted R2	.233*	.189*	.084‡	.083‡

^*p ≤ .001,

^†p ≤ .01,

^‡p < .05

Table III.

Effect of low birth weight (< 2.5 kg) on baseline growth in IG vs NIG (± s.e.).

Measures	IG < 2.5 kg (N =27)	IG ≥ 2.5 kg (N =85)	NIG ≥ 2.5 kg (N =65)
Baseline Height z-score (SD)	−1.19 (.15) ‡, *	−.70 (.10) *	.13 (.12)
Baseline Weight z-score	−1.80 (.20) ‡, *	−1.04(.12) *	−.09 (.13)
Baseline OFC z-score	−1.35 (.17) *	−.97 (.11) *	−.15 (.11)
Baseline Wt-for-Ht z-score	−1.12 (.23) ‡, *	−.56 (.13) *	.11 (.13)

^‡p ≤ .05 vs. IG ≥ 2.5 kg,

^*p < .001 vs NIG ≥ 2.5 kg

Equal variance was assumed unless Levine’s test for equality of variance was significant (< .05).

Post-Randomization Growth

CQS at 30 and 42 months improved for FCG and did not differ significantly from NIG (Table I). Children randomized to CAU were deinstitutionalized at the discretion of child protection officials and only 48% remained institutionalized at 42 and 35% at 54 months of age. Consequently, caregiving scores improved over time for CAU as well.

Growth was first examined from baseline to 42 months of age, 19 mo±6.7 (s.d.) months (range 9.1–31.2 mo, 93% ≥12 mo) after randomization. Random effects linear growth models (SAS PROC MIXED) with monthly measurements between baseline and 42 months of age in raw, non-standardized form indicated that children in FCG grew significantly faster in height and weight than CAU (Fig. 2a). Growth in head circumference did not differ between FCG and CAU. We also conducted a one-way within-subjects ANOVA within each group with the factor being time (baseline, 30 months, 42 months) and the dependent variables being measurements (Fig. 2b,c). When corrected for regression to the mean,⁴⁴ only height (+.50z) and weight z-scores (+.46z) in FCG and weight z-scores (+.08z) in CAU showed improvement.

Figure 2.

Growth in the foster care group (FCG) vs care as usual group (CAU) at baseline, 30 and 42 months of age.

a) Mean growth rates (s.e.) from random effects linear growth models for height (cm/mo), weight (kg/month) and OFC (cm/month) * p < .001. FCG 0.81 ± 0.02 cm.(s.e.), 0.26 ± 0.01 kg. per month, vs. CAU 0.71 ± 0.02 cm., 0.17 ± .01 kg.

b) Comparison of ν height and λ weight z-scores (s.e.) from baseline to 42 months of age. ANOVA FCG height [F(1, 50) = 100.6, p < .001, partial η² = .67] and weight z-scores [F(1, 50) = 94.0, p = < .001, partial η² = .65]. CAU weight z-scores [F(1. 50) = 10.3, p < .01, partial η² = .17]

At 42 months, height-z in FCG vs. CAU 0.06 ± .97 (s.d.) vs −0.62 ± 0.99, t(1,108) = −3.65, * p < .001 and weight-z −0.31 ± 1.05 vs −0.75 ± 1.17, t(1,108) = −1.98, † p = .05

c) Comparison of υ weight-for-height σ and ofc z-scores (s.e.).

To study the effect of intervention duration, we plotted growth at six-month intervals over the 18 months following randomization (Fig. 3). Measurements were included in the analyses if they were obtained within a ± 1-month window of the target intervals (6, 12, 18 months). FCG showed rapid increases in height and weight z-scores during the first 12 months while CAU showed no improvement. When corrected for regression to the mean, height (+.30z) and weight z-scores (+.48z) changed significantly. By 12 months 100% of FCG were in the normal range (≥ −2) for height, 90% for weight and 94% for weight-for-height. No significant change in any parameter occurred between 12 and 18 months post-placement.

Figure 3.

Growth in the foster care group (FCG) vs care as usual group (CAU) at baseline, 6, 12 and 18 months post-randomization.

a) Comparison of ν height and λ weight z-scores (s.e.m.) in CAU and FCG from baseline (n=51, 54) 6 ((n=46, 57), 12 (42, 51) and 18 (21, 50) months post-randomization. ANOVA FCG height [F(1, 43) = 57.32, p < .001, partial η² = .57] and weight z-scores [F(1, 43) = 63.65, p = < .001, partial η² = .60].

* p ≤ .001, † p <0.05 FCG vs. CAU at 12 and 18 months.

b) Comparison of υ weight-for-height and σ ofc z-scores (s.e.m.) from baseline to 18 post-randomization.

Factors Associated with Growth Recovery

We next examined factors influencing z-score changes (Δz) in FCG. To determine whether there were sensitive periods for catch-up growth, as previously reported for cognitive outcomes,¹⁸ attachment²⁴ and certain measures of electroencephalogram power and coherence¹⁵ in BEIP, five dichotomized ages of entry into foster care (12, 15, 18, 21, 24 months) were tested. Multiple regression models accounted for significant variance in Δz for all four measures (Table IV). Significant unique predictors of greater Δz included lower baseline z-scores (height, weight, OFC) and age of randomization <12 months (height, weight, OFC, weight-for-height). Higher post-placement CQS were also significant unique predictors of catch-up in height and weight, Individual components of the CQS quantifying caregiver detachment (height only) were negatively correlated while those quantifying positive regard for the child and sensitivity were positively correlated with Δz’s for height and weight (Table V).

Table IV.

a) Multiple regression of Δ z-scores on gender, ethnicity, birth weight, caregiving quality, age and baseline z-scores in FCG after randomization.

	ΔHeight z-score	ΔWeight z-score	ΔOFC z-score	ΔWt/Ht z-score
	β (SE β)	β (SE β)	β (SE β)	β (SE β)
Intercept	.04 (1.18)	.44 (1.39)	1.09 (1.12)	1.47 (1.70)
Gender (f = 0, m = 1)	−.01 (.20)	.39 (.23)	.24 (.18)	.55 (.28)
Ethnicity (Romanian = 0, other = 1)	−.32 (.20)	−.17 (.24)	−.15 (.19)	−.09 (.29)
Birth Weight (kg.)	.10 (.19)	.08 (.21)	.11 (.17)	−.20 (.24)
Post-Placement Caregiving (mean CQS at 30 & 42 months)	.62 (.19) †	.76 (.24) †	.05 (.18)	.56 (30)
Duration of Intervention (age at 42 month testing – placement age)	−.01 (.02)	−.01 (.03)	−.02 (.02)	−.03 (.03)
Age (< 12 mo = 0, ≥ 12 mo = 1)	−1.13 (.49) ‡	−1.79 (.57) †	−1.16 (.46)*	−2.29 (.71) †
Baseline z-score	−.36 (.14) ‡	−.37 (.12) †	−.29 (.09) †	−.26 (.14)
Adjusted R2	.271†	.391*	.359*	.295†

^*p ≤ .001,

^†p ≤ .01,

^‡p < .05,

Table V.

Pearson correlations between mean (30 and 42 mo) ORCE caregiver interaction scale ratings and Δ z-scores post-randomization in FCG.

ORCE Qualitative Ratings of Caregiver Interactions1	Δ z-scores Baseline to 42 Months (N = 53)
	Ht-z	Wt-z	OFC	Wt/Ht
Detachment 2 Caregiver is emotionally uninvolved, disengaged, and unaware of infant’s needs.	−.410†	−.149	−.016	−.009
Flat affect 2 Caregiver expresses no emotion or animation.	−.124	−.188	.05	−.123
Positive regard for the child Caregiver expresses positive feelings in interactions with the child.	.453*	.329‡	.125	.208
Sensitivity Caregiver responds to the infant’s social gestures and is attuned to the infant’s needs and moods.	.412†	.291‡	.054	.163
Stimulation of cognitive development Caregiver engages in activities that can facilitate the infant’s learning.	.201	.236	−.062	.175
Total Caregiving Quality Score CQS	.398†	.289‡	.016	.160

^*p ≤ .001,

^†p ≤ .01,

^‡p < .05

¹Scores range from 0 (not at all characteristic) to 4 (highly characteristic)

²Items that were reverse scored when calculating the total ORCE score

There were no significant gender differences in z-scores at 42 months or Δz’s baseline to 42 months and there were no significant differences between low (<2.5 kg) and normal (≥2.5 kg) birth weight infants in Δz’s baseline to 42 months (Table VI). While low birth weight infants tended to be smaller at 42 months of age, the difference was significant for head size alone.

Table VI.

Effect of low birth weight (<2.5 kg) and gender on catch-up growth and cognitive measures following randomization to FCG, mean z-scores (± s.e.).

Measures	Male (N = 31)	Female (N = 32)	< 2.5 kg (N = 17)	≥ 2.5 kg (N = 40)
42 month Height z-score	−.10 (.16)	.22 (.20)	−.26 (.22)	.19 (.17)
42 Month Weight z-score	−.35 (.20)	−.28 (.24)	−.59 (.21)	−.25 (.22)
42 Month OFC z-score	−1.28 (.21)	−1.12 (.21)	−1.67 (.25)	−.93 (.17) *
42 Month Wt-for-Ht z-score	−.54 (.22)	−.63 (.29)	−.70 (.27)	−.61 (.26)
Δht-z bl-42 mo	.92 (.16)	.94 (.13)	.87 (.18)	.99 (.14)
Δwt-z bl-42 mo	1.15 (23)	1.04 (.15)	1.06 (18)	1.10 (.19)
Δofc-z bl-42 mo	.06 (.14)	−.13 (.15)	−.11 (.15)	.02 (.14)
Δw/ht-z bl-42 mo	.41 (.23)	.23 (.21)	.30 (.21)	.29 (.21)
Baseline ORCE	2.23 (.11)	2.35 (.10)	2.34 (.15)	2.23 (.10)
Mean ORCE at 30 and 42 mo.	2.70 (.07)	2.81 (.10)	2.82 (.12)	2.73 (.08)

^*p < .05

Equal variance was assumed unless Levine’s test for equality of variance was significant (< .05).

Growth and Cognitive Improvement in the Foster Care Group

Reestablishment of normal kinetics in the growth hormone-IGF-1 axis (GH-GF-1 axis) has been shown to be a factor in growth recovery in neglected children placed in more nurturing environments. Based on this fact and on reports of improved stature correlating with cognitive gains in stunted, cognitively impaired children treated with growth hormone ^45–47, we investigated whether catch-up growth in FCG could predict cognitive abilities at 42 months, and if so, whether the effects would persist at 54 months. Included were child characteristics, CQS in foster care, baseline OFC, Δz-scores between baseline and 42 months in height, weight and OFC and the binary age variable (24 months) previously reported to define a sensitive period for cognitive development in this cohort.¹⁸ Multiple regression models accounted for significant variance in both cognitive measures (Table VII). Baseline DQ and Δz-height were both significant unique predictors positively related to DQ-42 and Full IQ-54. Birth weight and female gender were significant unique predictors positively related to Full IQ-54. As predicted by the risk factors of low birth weight and delayed placement into foster care, ¹⁸ children with birth weights ≥2.5 kg placed in foster care prior to 24 months of age had a mean IQ score 25 points higher than LBW infants placed after 24 months of age (Fig. 4).

Table VII.

Multiple regression of cognitive variables on gender, ethnicity, birth weight (kg.), baseline developmental quotient (DQ), post-placement caregiving, age, and catch-up growth in institutionalized children randomized to foster care (FCG).

	DQ-42 mo n = 47	Δ-DQ Bl-42 mo n = 47	Full Scale IQ-54 mo n = 45	Δ-DQ/IQ Bl-54 mo n = 45	Performance IQ-54 mo n = 45	Verbal IQ-54 mo n = 45
Variable	β (SE β)	β (SE β)	β (SE β)	β (SE β)	β (SE β)	β (SE β)
Intercept	41.25 (17.31)	41.25 (17.31)	15.38 (27.18)	15.38 (27.18)	29.49 (25.67)	19.11 (27.00)
Gender (f = 0, m = 1)	−4.60 (3.24)	−4.60 (3.24)	−8.09 (5.09)	−8.09 (5.09)	−8.61 (4.81)	−6.06 (5.05)
Ethnicity (Romanian = 0, other = 1)	−5.00 (3.26)	−5.00 (3.26)	−2.04 (5.07)	−2.04 (5.07)	−7.45 (4.79)	2.80 (5.04)
Birth Weight (kg.)	−.25 (2.70)	−.25 (2.70)	9.22‡ (4.33)	9.22‡ (4.33)	9.49‡ (4.09)	7.04 (4.30)
Baseline DQ	.45† (.13)	−.55* (.13)	.46‡ (.20)	−.54* (.20)	.38 § (.19)	.45‡ (.20)
Post-Placement Caregiving (mean ORCE at 30 & 42months)	5.49 (3.47)	5.49 (3.47)	2.48 (5.49)	2.48 (5.49)	.28 (5.18)	4.47 (5.45)
Age < 24 mo. = 0, ≥ 24 mo. = 1	−3.85 (3.89)	−3.85 (3.89)	−2.73 (5.96)	−2.73 (5.96)	1.68 (5.64)	−5.14 (5.92)
Δht-z bl-42 mo	8.14‡ (3.61)	8.14‡ (3.61)	9.28 § (4.71)	9.28 § (4.71)	3.31 (4.45)	12.36‡ (4.68)
Δwt-z bl-42 mo	−4.19 (2.44)	−4.19 (2.44)	−3.35 (3.83)	−3.35 (3.83)	1.78 (3.62)	−7.32 (3.80)
Δofc-z bl-42 mo	−.78 (2.27)	−.78 (2.27)	−2.50 (3.40)	−2.50 (3.40)	−4.32 (3.21)	−1.61 (3.34)
Adjusted R2	.545*	.566*	.404‡	.418‡	.379‡	.403‡

^*P ≤ .001,

^†P < .01,

^‡P < .05,

^§P < .06

Developmental quotient (DQ) at 42 months, ΔDQ at 42 months of age [DQ at 42 months – baseline (BL) DQ], WPPSI full scale, performance and verbal IQ scores at 55 months and the ΔDQ/IQ between baseline and 55 months of age. SE = standard error of estimate

Figure 4.

Full Scale IQ ± SE at 54 months vs. age of randomization and birthweight.

ν ≥ 2.5 kg, λ < 2.5 kg, * p < 0.001 vs. children ≥ 2.5 kg randomized at < 24 months [66.6 ± 6.2(s.e.), vs. 91.1±3.9 t(21) = 3.84]

Discussion

Children raised in institutions had globally suppressed growth followed by recovery of height and weight but not OFC within those removed and placed in foster care. These observations mirror the report of suppressed growth within institutional care settings and recovery within adoptive families reported in a meta-analysis by Van Ijzendoorn and colleagues.¹¹ Both the current data and the meta-analysis showed that height catch-up improved if placement occurred prior to 12 months of age. Data from BEIP confirmed the same sensitive period for weight, OFC and weight-for-height. In FCG, catch-up growth for height and weight were robust and essentially complete by 12 month after randomization when both height and weight z-scores were close to 0.

Though growth has shown improvement within cohorts of neglected and or abused children once removed from an adverse environment,^{9, 11, 48, 49} the contribution of individual child-caregiver interaction to catch-up growth has never been directly explored. The BEIP is the first to evaluate growth in children living under adverse social circumstances in relation to standardized measures of individual caregiving environments. The data indicate that caregiving quality was a significant independent predictor of catch-up growth in height and weight. Components of the CQS that were positively correlated with catch-up included sensitivity (child-centered, contingent responses) and positive regard for the child (acceptance, respect and warmth including expressions of physical affection).

Following the initial report of Talbot and colleagues ⁵⁰ investigators have described four syndromes of impaired growth associated with adversity based on several factors including: age, nutrition, behavioral or emotional co-morbidities and status of the GH-IGF-1 axis.^2–6 All share the diagnostic features of suppressed growth within the context of adversity followed by catchup growth after improvement in caregiving environment; a central finding observed in children in FCG. Caloric deprivation has been reported to play a central role in growth failure during infancy (Type I), while changes in the GH-IGF-1 axis (Types IIA, IIB and III) becomes more important in growth suppression and recovery in children beyond the first 18–24 months of life.^2–6 Both nutrition and production of endogenous growth factors are likely to be affected by an institutional environment.

Nutritional requirements in children vary depending on growth rates and whether preexisting deficits exist. During the rapid growth phase between birth and 18 months the effects of even modest nutritional deficits become magnified. Both low birthweight infants and children with orofacial malformations or neuromotor problems are overrepresented within institutional care settings and may have difficulty obtaining and/or consuming sufficient calories to grow.^{10, 12, 22, 51} With the time and fiscal constraints experienced by virtually all orphanages worldwide, it is highly unlikely that the nutritional needs of individual children can be accommodated within an environment where dietary plans and feeding protocols are strictly regimented and caregiver actions are based on efficiency and expediency rather than being responsive to child-based cues.⁵²

Considering that neglected infants are highly susceptible to insufficient intake, it is not surprising that malnutrition is felt to be the principal cause of deprivation-associated growth failure within this age group.^{2, 53} Two observations in BEIP including, lower CQS during infancy, and that wasting was significantly more common in younger children underscore the importance of malnutrition as a major determinant in psychosocial growth failure during infancy. Outside of the first years of life, there is less evidence that caloric deprivation is a primary factor in the etiology of growth failure in neglected and/or abused children as weight-for-height has been reported to be essentially normal.^{2, 3, 6} Consistent with past studies, weight-for-height in the IG at baseline was greater in older children and only 6% of children > 24 months were wasted vs. 25% in those < 12 months.

In older children, alterations in the GH-IGF-1 axis have been documented to play a key role in growth failure and catch-up.^{2–6, 48} In a cohort of post-institutionalized Eastern European children of similar age to those in BEIP (mean 20.4 months, range 7.3–59.9), levels of IGF Binding Protein-3 shortly after placement was an independent predictor of height z-score.¹² As would be expected with abnormal GH-IGF-1 axis function, height but not weight, head circumference or weight-for-height at baseline in BEIP was lower in older institutionalized children (IG).

It was impossible for us to directly measure serum growth factors given constraints imposed on us by local authorities and IRBs and the labor intensive process of nutritional evaluation could not be conducted with resources available. However, auxology in BEIP is consistent with previous work in this area and implicates both nutrition and alterations in the GH-IGF-1 axis in growth suppression. The relative importance of these two factors and the consequent clinical presentations almost certainly relate to a child’s age with malnutrition contributing more during periods of rapid growth and utter dependency and depression of the GH-IGF-1 axis becoming more important as growth rates slow, children are more able to regulate their own dietary intake and linear growth becomes more dependent on growth hormone. ^{54, 55}

The observation that catch-up growth in height was the only significant independent auxologic predictor of cognitive abilities at 42 and 54 months suggests that the GH-IGF-1 axis may also play a role in cognitive recovery. The role of this complex system in cognitive development is supported by substantial experimental,^{56, 57} and clinical evidence. In normal 8- to 9-year-old children IGF-1 levels were shown to be positively related to IQ.⁵⁸ Children with 18q deletions,⁴⁵ Prader-Willi Syndrome⁴⁶, or born SGA,⁴⁷ conditions characterized by both short stature and cognitive delays, have shown significant improvement in height, IQ and brain structure following treatment with GH. Finally, children with defects in the GH receptor, have IQs and brain structural abnormalities that differ depending on which exon contains the point mutation or deletion.⁵⁹

Irrespective of whether there are direct causative relations, improvement in stature proves not only to be a useful biologic measure of caregiving environment but an informative indicator of cognitive improvement in at-risk children. As countries heed the call to eliminate institutions by developing kinship and foster care, post-placement growth could be used as a cost-effective marker of caregiving quality and child well being pending sufficient staffing and funding of more comprehensive social services programs.

From this study, several significant themes emerge with relevance to policies for care of institutionalized children. Infants who are LBW are particularly vulnerable to the effects of social deprivation and should be the first triaged to family care. The unique nutritional needs of LBW infants are unlikely to be appreciated and even if they are, dependence on conformity in orphanages makes it unlikely that specific nutritional interventions needed to optimize growth would be offered for growth retarded or premature infants. Growth at baseline in our institutionalized LBW infants was particularly impaired, consistent with the global growth failure described in LBW infants who experienced social deprivation in early life.^{12, 60–62} They were also at higher risk for cognitive compromise and smaller head size at 42 months, findings that have also been observed in LBW infants exposed to social deprivation.^61–63 Second, the sensitive period for growth recovery (< 12 months) is even earlier than the sensitive period (< 24 months) described in BEIP for cognitive recovery,¹⁸ attachment behavior²⁴ and improvement in the electroencephalogram¹⁵ strengthening the argument for placement within family care as early as possible. Not only will delayed placement affect growth in LBW as it does for all infants but the combination of LBW plus prolonged duration of institutional care > 24 months ¹⁸ resulted in severe cognitive compromise. Finally, though family is clearly preferable to institutional care, establishing higher quality child-caregiver interactions with appropriate screening, training in emotional engagement and contingent caregiving and monitoring insures the best outcomes in terms of growth.

The significance of these findings extends beyond the millions of institutionalized children worldwide to hundreds of millions of impoverished children who are stunted and/or do not meet their developmental potential. The interdependence of nutrition and social environment on child outcomes has recently received attention in regards to achieving UN Millennium Development Goals.⁶⁴ This study of growth in institutionalized children adds strong experimental support to the conclusion of Black and colleagues, that strategies that fail to address nurture along with health and nutrition will likely fail to achieve significant improvements in overall child well-being.⁶⁴ Psychosocial deprivation within any caregiving environment during early life is as detrimental as malnutrition and must be viewed with as much concern as any severely debilitating childhood disease.

Abbreviations

BEIP

The Bucharest Early Intervention Project

CAU

Care as Usual Group

CQS

Caregiving Quality Score

DQ

Bayley II Developmental Quotient

Δz

Change in z-scores

FCG

Foster Care Group

GH

Growth Hormone

IG

Institutionalized Group at Baseline

IGF-1

Insulin-Like Growth Factor 1

LBW

Low Birth Weight

NIG

Never Institutionalized Group

OFC

Occipital-frontal circumference

ORCE

Observational Record of the Caregiving Environment

SGA

Small for Gestational Age

WPPSI

Wechsler Preschool Primary Scale of Intelligence