The interaction of prematurity with genetic and environmental influences on cognitive development in twins

The interaction of prematurity with genetic and environmental influences on cognitive development in twins Gesina Koeppen-Schomerus, BSc, Thalia C. Eley, PhD, Dieter Wolke, PhD, Paul Gringras, MRCPCH, and Robert Plomin, PhD Objective: To investigate how the degree of prematurity interacts with genetic and environmental influences in their effect on verbal and nonverbal cognitive development. Study design: The target sample consisted of more than 2000 pairs of twins born in England and Wales in 1994. At 24 months, measures of verbal and non-verbal cognitive development were obtained from the twins’ parents. The sample was divided into 3 groups according to degree of prematurity: very preterm or high-risk (<32 weeks), moderately preterm or medium-risk (32-33 weeks), and mildly preterm/term or low-risk (>34 weeks). Quantitative genetic analyses were used to assess the contributions of genetic and environmental influences on vocabulary and cognitive development. Results: The results indicated gene-environment interactions. For the high-risk group, genetic effects on both verbal and non-verbal cognitive ability were completely overshadowed by shared environmental factors, whereas for both medium- and low-risk groups, additive genetic effects explained 18% to 33% of the variance. Conclusions: Our findings indicate that genetic factors are not responsible for cognitive outcomes of very preterm infants and suggest that early environmental influences appear to affect verbal and non-verbal cognitive development at 2 years of age. (J Pediatr 2000;137:527-33)

neonatal complications and are exposed to highly intensive and prolonged neonatal intensive care, and most often have long-term cognitive impairments.2-5 Although the assessment of factors such as brain injury6 have contributed to the understanding of the pathogenic pathways associated with severe cognitive deficits in preterm infants, the mediating factors that predispose preterm children to cognitive impairments remain poorly understood.7 It is unknown whether genetic factors play a major role in the development of individual differences in cognitive development in preterm infants. Alternatively, differences in cognitive outcome in preterm children might be due to environmental factors such as prenatal environment,8,9 perinatal or neonatal experiences,10,11 or post-discharge family environment.12 MCDI

Premature birth is associated with a range of neonatal complications and increased risk of adverse developmental outcome in childhood.1 Within the

group of prematurely born infants, cognitive outcome is variable. Those born at very low birth weight and very low gestation experience a large number of

From Social, Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, London, United Kingdom; Department of Psychology, University of Hertfordshire, Hatfield, Hertfordshire, United Kingdom; and Multiple Births Foundation, Queen Charlotte’s and Chelsea Hospital, London, United Kingdom. This research is part of the Twins’ Early Development Study (TEDS) which is supported by a program grant from the Medical Research Council (UK).

Submitted for publication Oct 7, 1999; revisions received Feb 17, 2000, and Apr 17, 2000; accepted May 2, 2000. Reprints not available from author. Copyright © 2000 by Mosby, Inc 0022-3476/2000/$12.00 + 0 9/21/108445 doi:10.1067/mpd.2000.108445

MacArthur Communicative Development Inventory PARCA Parent Report of Children’s Cognitive Abilities TEDS Twins’ Early Development Study

Genetically sensitive designs such as the twin method can disentangle childdriven genetic factors and shared environmental factors such as neonatal intensive care to the etiology of individual differences in cognitive and language development for premature children.13 We investigated these questions by studying high-risk (gestation <32 weeks), moderate-risk (gestation 32-33 weeks), and low-risk infants (gestation ≥34 weeks) within a large representative sample of twins. 527

KOEPPEN-SCHOMERUS ET AL

THE JOURNAL OF PEDIATRICS OCTOBER 2000

Table I. Sample characteristics and demographic details for high-risk, moderate-risk, and low-risk groups

High-risk group (<32 wk) Gestation (wk) Mean SD Birth weight (g) Mean SD Duration of hospital treatment (d) Mean SD Maternal age (y) Mean SD Sex of child Male (%) Female (%) Special care after birth Yes (%) No (%) Maternal education University degree (%) Completed further education* (%) Completed mandatory education† (%) Without formal qualification (%)

Moderate-risk group (32-33 wk)

29.5 1.6

32.5 0.5

Low-risk group (>33 wk) 37.1 1.7

1369 337

1859 468

2592 456

57 46

26 13

8 6

31 5.3

32 5.1

32 4.8

51.8 48.2

50.8 49.2

47.6 52.4

95 5

35 65

14.1 16.9 59.8 9.2

16.1 17.0 57.9 9.0

100 0 11.3 23.3 57.0 8.4

*Obtained A-levels or “Higher National Diploma/Certificate” (UK qualifications) †Obtained UK secondary school qualifications “GCSEs” or “CSEs.”

METHODS The sampling frame for this study, the Twins Early Development Study, consisted of all twins born in England and Wales in 1994.14 Unlike previous reports,14-16 this study included twins who were born very premature. The initial target sample consisted of 3452 families who returned both background information on the twins and their families and test booklets for each twin, which included the cognitive measure and the vocabulary measure described below. At each stage of the study (ie, when background information was obtained, as well as when test booklets were returned), parents signed a consent form agreeing to participate in the TEDS; they were also aware that they could withdraw from the study at any time. A total of 1229 528

pairs were excluded from the analyses for the following reasons: missing information on the co-twin (n = 5), sex unknown (n = 5), high maternal alcohol consumption during pregnancy (n = 12), debilitating medical problems (n = 29), hearing problems (n = 12), missing information on gestational age or zygosity and opposite sex twin pairs (n = 1166). The potential target sample for the analyses thus consisted of 2223 twin pairs (1134 monozygotic pairs and 1089 dizygotic same-sex pairs). Twin zygosity was assigned by using parent questionnaire ratings of twins’ physical similarity. A previous report in which a zygosity questionnaire was used showed accuracy between 93% and 98%.17 In a separate analysis on the instrument used in our study, zygosity was correctly assigned by parent ratings in 94.7% of cases as vali-

dated against zygosity assigned by identity of polymorphic DNA markers.18 The sample was divided into 3 groups (Table I) according to degree of prematurity (gestation was based on parent reports): 1. Very preterm or high-risk group born between 25 and 31 weeks’ gestation (5.0% of the target sample; mean birth weight, 1369 g; interquartile range, 457 g) 2. Preterm or medium-risk group born at 32 or 33 weeks’ gestation (8.6% of the target sample; mean birth weight, 1858 g; interquartile range, 439 g) 3. Moderately preterm/term or lowrisk group, which included all infants with a gestational age of 34 weeks’ or above (86.4% of the target sample; mean birth weight, 2591 g; interquartile range, 609 g).

KOEPPEN-SCHOMERUS ET AL

THE JOURNAL OF PEDIATRICS VOLUME 137, NUMBER 4 We compared the sample after exclusions for both measures, the Parent Report of Children’s Cognitive Abilities and MacArthur Communicative Development Inventory, and found no significant differences from the original target sample in terms of basic demographic descriptors.

Procedures and Definitions VERBAL COGNITIVE DEVELOPMENT (MCDI). Verbal development was assessed in relation to vocabulary production at 2 years of age by using an adapted version of the MCDI,19,20 which is based on parents’ reports of their children’s word use. The MCDI has excellent internal consistency, test-retest reliability, and concurrent validity.19 MEASURE-SPECIFIC EXCLUSIONS (MCDI). For analyses of the verbal cognitive measure (MCDI), an additional 163 pairs of twins were excluded from the analyses because English was not the first language, MCDI total scores were missing, the measure was completed too soon or too late (ie, 2 months before or 6 months after the twins’ second birthday), or there was conflicting information on language production. The final sample for the MCDI consisted of 2060 twin pairs (1049 monozygotic and 1011 dizygotic same-sex). NON-VERBAL COGNITIVE DEVELOP(PARCA). Non-verbal cognitive development was assessed at 2 years of age by using the PARCA.21 The PARCA consists of parent-administered items, that is, non-verbal cognitive tasks such as design copying, item matching, block building, and imitative action, as well as parent report items on their children’s abilities (eg, “Does your child recognize himself/herself when looking in the mirror?”). The PARCA score was standardized to a mean of zero and a standard deviation of 1. MENT

MEASURE-SPECIFIC EXCLUSIONS (PARCA). For analyses of the non-verbal

Fig 1. Mean standardized scores for total sample at 2 years for language development (MCDI) and cognitive performance (PARCA) by gestational risk category.

cognitive measure, 243 pairs were excluded because of missing total scores or completion of the measure too early or too late. The total PARCA sample consisted of 1980 twin pairs (1001 monozygotic and 979 dizygotic same-sex).

Statistical Analyses The study used the twin design, which takes advantage of the differing levels of genetic relatedness between monozygotic twin pairs who are genetically identical and dizygotic twin pairs who are, on average, 50% similar genetically. This difference is used to assess the contributions of genetic and environmental factors to the individual differences for the relevant phenotype under investigation. The variance in the phenotype is attributed to 3 different latent factors: additive genetic variance (a2), shared environmental factors (c2), and non-shared environmental factors (e2), which also include measurement error. Within-pair similarity for the phenotype is assumed to be due to genetic factors plus common or shared environment factors that make members of a family similar. Non-shared or unique environment is a residual term that includes environmental factors that make members of a family different from one another and measurement error. By comparing the within-pair correlations

for monozygotic and dizygotic twins, estimates for the contribution of additive genes, shared environment, and unique environment to the variance in the phenotype under investigation can be obtained. The effects of age and sex were regressed out, as is standard practice in twin research, because these variables can inflate twin similarity. To obtain estimates of genetic and environmental effects on both developmental measures, standard maximum likelihood model fitting was applied to the variance-covariance matrices by using the structural equation modeling package Mx.22 A univariate model was fitted to the observed data for both verbal and non-verbal cognitive development measures assessed at 2 years.

RESULTS Descriptive Statistics Fig 1 compares the mean standardized scores for both verbal and nonverbal cognitive measures for high-, medium-, and low-risk groups. The data show a linear relationship between degree of prematurity and mean performance on both outcome measures. These findings are in line with previous reports, which have consistently demonstrated delayed performance for very preterm children. However, it 529

KOEPPEN-SCHOMERUS ET AL

THE JOURNAL OF PEDIATRICS OCTOBER 2000

Fig 2. Comparison of monozygotic (MZ) twin correlations with dizygotic (DZ) twin correlations at different levels of gestational risk for measures of verbal (MCDI) and non-verbal (PARCA) cognitive development.

should be noted that these mean differences only account for a small amount of variance and that the range of individual differences in performance within the high- and moderate-risk groups is just as great as in the low-risk group.

Univariate Quantitative Genetic Analyses Fig 2 shows the twin intraclass correlations for verbal (PARCA) and nonverbal (MCDI) cognitive development for monozygotic and dizygotic samesex twin pairs as a function of gestational risk. Within the high-risk group, the twin correlations for both outcome measures were similar for both monozygotic and dizygotic same-sex twins, indicating no genetic influence and substantial shared environmental influence. For the moderate- and low-risk groups, the differences in correlations between monozygotic and dizygotic same-sex twins were more pronounced, implying a greater contribution of genetic factors and less shared environmental influence with increasing gestational maturity. These interpretations, gleaned from examination of the twin correlations, were confirmed and 530

shown to be significant by means of structural equation modeling. The following 4 models were tested for their relative goodness of fit for both cognitive measures: Model 1: a free, unconstrained model that assumes different parameter estimates (a2, c2, and e2) for each risk group individually Model 2: a more constrained model that compares the high-risk group with both moderate-risk and lowrisk groups taken together Model 3: like model 2 except that it compares the high-risk plus moderate-risk groups with the lowrisk group Model 4: a fully constrained model assuming equal parameter estimates across all groups of different gestational risk

Model-Fitting Results for Verbal Cognitive Development (MCDI) The patterns of twin correlations were formally tested for the MCDI by applying structural equation model fitting. As indicated in Table II, model 1

yielded the best fit for the data on vocabulary development. This model assumes distinct genetic and environmental parameter estimates across the 3 gestational risk groups. Fixing parameters to be equal across groups in models 2 to 4 did not improve the fit. For instance, when comparing only 2 groups, that is, by testing whether parameter estimates for the moderateand high-risk children are similar to each other and different from those in the group of low-risk children (model 2), the fit significantly worsened with changes in χ2 of 50.989. This was also the case for model 3 comparing moderate- and low-risk children taken together and the high-risk children, which also yielded a significant change in χ2 at 60.145. A poorer fit was further obtained for the fully constrained model (model 4), which assumed parameter estimates to be equal for the 3 risk groups, resulting in a significant change in χ2 of 65.788. Thus model 1 provides distinct parameter estimates for each risk group for additive genes (a2), shared environment (c2), and unique environmental influences (e2). These estimates (Fig 3) show a striking pattern in that for the high-risk group most of the variance is explained by shared environmental factors (84%) with negligible and nonsignificant additive genetic effects (9%). For both moderate- and lowrisk groups, genetic effects are significant (33% and 22%, respectively), and shared environment is lower (65% and 73%) than in the high-risk group. For the moderate- and low-risk groups, heritability and shared environmental influences are comparable to the parameter estimates previously reported for the TEDS sample (a2 = 25%, c2 = 69%, e2 = 6%).14

Model-Fitting Results for NonVerbal Cognitive Development (PARCA) Model-fitting results for non-verbal cognitive development are similar to those for the MCDI in suggesting a

KOEPPEN-SCHOMERUS ET AL

THE JOURNAL OF PEDIATRICS VOLUME 137, NUMBER 4

Fig 3. Percentage of variance explained by genes, shared environment, and non-shared environment in high-, moderate-, and low-risk gestational groups for verbal (MCDI) and non-verbal (PARCA) cognitive development.

Table II. Model-fitting results for verbal and non-verbal cognitive development at 2 years as measured, respectively, by the MacArthur Communicative Development Inventory and the PARCA

MCDI Model 1 Model 2 Model 3 Model 4 PARCA Model 1 Model 2 Model 3 Model 4

χ2

df

P value

RMSEA*

AIC*

∆ χ2

∆ df

P value

8.72 59.71 68.87 74.51

9 12 12 15

.46 .00 .00 .00

.005 .070 .118 .123

–9.28 47.71 56.87 68.58

— 50.99 60.15 65.79

— 3 3 6

— <.001 <.001 <.001

11.54 12.69 13.38 16.53

9 12 12 15

.24 .05 .04 .00

.000 .000 .001 .001

–6.47 0.69 1.38 10.53

— 1.15 1.84 4.99

— 3 3 6

— NS NS NS

Model 1 allows all gestational age/risk groups to differ from one another (ie, high vs moderate vs low). Model 2 opposes the high-risk group (<32 weeks) to low- and medium-risk groups taken together. Model 3 is the same as model 2 but contrasts both high- and medium-risk groups taken together against the low-risk group. Model 4 assumes all groups to be equal in terms of gestational risk. RMSEA, Root Mean Squared Error Approximation; AIC, Akaike’s Information Criterion; NS, not significant. *Statistical fit indices.

gene-environment interaction in which genetic influence is negligible for the high-risk gestational group. Again, the fully constrained model (model 1) fits best (see Table II). However, model 1 did not fit significantly better than models 2, 3, and 4, suggesting that genetic and environmental parameter estimates do not differ significantly between low-, moderate-, and high-risk

groups for PARCA. When comparing model 1 with models 2 to 4, the changes in χ2 were all non-significant, suggesting no improvement in fit.

DISCUSSION The results of this study are an example of significant genotype-

environment interactions in humans. Gene-environment interaction refers to genetic differences in sensitivity or susceptibility to environments.13 In this study, gene-environment interaction refers to the interaction between the environmental trauma of prematurity and the magnitude of genetic effects. We found that individual differences in 2-year-old cognitive outcomes of very 531

KOEPPEN-SCHOMERUS ET AL

premature birth are not due to genetic factors. This finding implies a low likelihood of finding genetic markers that are involved in predicting cognitive outcome in very preterm infants. Instead, shared environmental factors have the greatest impact on cognitive scores in very premature infants, indicating the special importance of nurture and medical care in this group at high risk for cognitive impairment. Treatment regimens for prematurity would be read in the twin design as shared environment for premature twins because both members of a twin pair share this treatment. Shared environment can pertain to prenatal, perinatal, neonatal, or caretaking environment after discharge. Prenatal, perinatal, and neonatal complications such as cerebral hemorrhage, periventricular leukomalacia, respiratory distress syndrome, or poor catch-up growth constitute such environmental factors. We speculate that such complications arise because of prematurity and early vulnerability and may be partly related to iatrogenic complications. It is also possible that differences in care-taking environment might have overwhelmed genetic influences. For instance, interventions such as stimulation programs for high-risk preterm infants after discharge from the hospital aim to improve the shared environment. However, efficacy evaluations have shown few long-term beneficial effects of these interventions in very preterm infants, although benefits in preterm infants of greater gestational age have been reported.23-25 In contrast, there is evidence that neonatal complications such as respiratory problems may be partly related to iatrogenic complications of too-intensive neonatal treatment such as overuse of intubation and long duration of ventilation or parenteral nutrition.11,26-28 These factors can in turn have adverse effects on long-term cognitive development.29,30 Possible changes in the practice of neonatal care including gentler, individualized care have demonstrated 532

THE JOURNAL OF PEDIATRICS OCTOBER 2000 short-term benefits for preventing respiratory and other neonatal complications31 and led to improvements in early cognitive development.32 However, samples studied to date have been small and need replication. Our results show the great impact of shared environment within the highrisk group. However, we are unable to determine the specific mechanisms responsible for this shared environment, which constitutes factors such as prenatal environment, neonatal complications and treatment, and differences in post-discharge parental care. Further investigation of neonatal treatment experiences of very preterm twins and their relation to cognitive outcome is needed to distinguish among these alternatives. It is also necessary to consider possible influences of neonatal problems that are specific to twin type (eg, intrauterine growth discordance or twin-to-twin transfusion syndrome). We are currently in the process of collecting neonatal details from the twins’ hospital records and aim to implement those in future studies to shed more light on the nature of neonatal complications in very preterm twins and their impact on development. At present, we can only speculate on the reasons and require more detailed information on discordance or similarity of complications and treatment experienced within twin pairs. Limitations of this study include the issue of statistical power. Adequate power to detect significant differences in heritability indicative of gene-environment interaction requires very large samples. Although our sample was large enough to detect a significant geneenvironment interaction, these findings need to be confirmed in other samples. We are collecting similar data on the TEDS 1995 cohort, which should allow us to replicate the study results. A further limitation is that the outcome measures within this sample are based on parent reports and are limited to cognitive performance at 2 years of age. However, similar data on ver-

bal and non-verbal cognitive measures are being obtained at 3 and 4 years of age to explore how the pattern of genetic and environmental influences will unfold in relation to prematurity. National statistics indicate that death rates differ between the risk groups defined according to gestational age. This study includes only those twins who are reared by their parents and thus only concerns the survivors. Certainly, within a non-genetic study that includes data on the full range of outcomes (eg, deaths, stillbirths, live births, congenital abnormalities), adverse outcomes should be documented to indicate the range of deficits. If prematurity and related environmental factors loosen their grip on development, heritability will increase and shared environment will decrease in importance for very premature children. In contrast, little relative change of heritability in comparison with the later gestation groups would indicate a continued impact of the environmental trauma of prematurity. These findings represent a first step toward widening our understanding of the predisposing factors (ie, nature, nurture, and medical care) involved in shaping the psychologic development of very preterm infants. Once the mechanisms are better understood, they will provide the basis for developing preventative strategies, which could lead to enhanced development of infants born very premature.

REFERENCES 1. Paneth NS. The problem of low birth weight. Future Child 1995;5:19-34. 2. Wolke D, Meyer R. Cognitive status, language attainment, and pre-reading skills of 6-year-old very preterm children and their peers: the Bavarian Longitudinal Study. Dev Med Child Neurol 1999;41:94-109. 3. Botting N, Powls A, Cooke RW, Marlow N. Cognitive and educational outcome of very-low-birthweight children in early adolescence. Dev Med Child Neurol 1998;40:652-60. 4. Resnick MB, Gomatam SV, Carter

KOEPPEN-SCHOMERUS ET AL

THE JOURNAL OF PEDIATRICS VOLUME 137, NUMBER 4

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

RL, Ariet M, Roth J, Kilgore KL, et al. Educational disabilities of neonatal intensive care graduates. Pediatrics 1998;102:308-14. Taylor HG, Klein N, Schatschneider C, Hack M. Predictors of early school age outcomes in very low birth weight children. J Dev Behav Pediatr 1998; 19:235-43. Whitaker AH, Feldman JF, Van Rossem R, Schonfeld IS, Pinto-Martin JA, Torre C, et al. Neonatal cranial ultrasound abnormalities in low birth weight infants: relation to cognitive outcomes at six years of age. Pediatrics 1996;98:719-29. Wolke D. Language problems in neonatal at risk children: towards an understanding of developmental mechanisms [invited comment]. Acta Paediatr 1999;88:488-90. Brooks AA, Johnson MR, Steer PJ, Pawson ME, Abdalla HI. Birth weight: Nature or nurture? Early Hum Dev 1995;42:29-35. Sebire NJ, D’Ercole C, Hughes K, Rennie J, Nicolaides KH. Dichorionic twins discordant for intrauterine growth retardation. Arch Dis Child Fetal Neonatal Ed 1997;77:F235-6. Laucht M, Esser G, Schmidt MH. Developmental outcome of infants born with biological and psychosocial risks. J Child Psychol Psychiatry 1997;38: 843-53. Silverman WA. Overtreatment of neonates? A personal retrospective. Pediatrics 1992;90:971-6. Bendersky M, Lewis M. Environmental risk, biological risk and developmental outcome. Dev Psychol 1994;30: 484-5. Plomin R, DeFries JC, Rutter M. Behavioral genetics. 3rd ed. Hillsdale (NJ): Lawrence Erlbaum Associates; 1997. Dale PS, Simonoff E, Bishop DV, Eley

15.

16.

17.

18.

19.

20.

21.

22.

23.

TC, Oliver B, Price TS, et al. Genetic influence on language delay in twoyear-old children. Nat Neurosci 1998;1:324-8. Eley TC, Bishop DVM, Dale PS, Oliver B, Petrill SA, Price TS, et al. Genetic and environmental origins of verbal and performance components of cognitive delay in two-year-olds. Dev Psychol 1999;35:1122-31. Price TS, Eley TC, Plomin R. Genetic and environmental covariation between verbal and non-verbal cognitive development in infancy. Child Dev 2000;71(4). Goldsmith, HH. A zygosity questionnaire for young twins: a research note. Behav Genet 1991;21:257-69. Price TS, Freeman B, Plomin R. Infant zygosity can be assigned by parental report questionnaire data. Twin Research. In press. Fenson L, Dale PS, Reznick JS, Bates E, Thal DJ, Pethick SJ. Variability in early communicative development. Monogr Soc Res Child Dev 1994;59:1-185. Dale PS. Parent report assessment of language and communication. In: Cole KN, Dale PS, editors. Assessment of communication and language. Communication and language intervention series. Vol 6. Baltimore (MD): Paul H. Brookes Publishing Co; 1996. p. 161-82. Saudino KJ, Dale PS, Oliver B, Petrill SA, Richardson V, Rutter M, et al. The validity of parent-based assessment of the cognitive abilities of 2-year-olds. Br J Dev Psychol 1998;16:349-63. Neale MC. Mx: Statistical modeling. 4th ed. Richmond (VA): Department of Psychology; 1997. Avon Premature Infant Project. Randomised trial of parental support for families with very preterm children. Arch Dis Child Fetal Neonatal Ed 1998;79:F4-11.

24. Baumeister AA, Bacharach VR. A critical analysis of the infant health and development program. Intelligence 1996;23:79-104. 25. McCarton CM, Brooks-Gunn J, Wallace IF, Bauer CR, Bennett FC, Bernbaum JC, et al. Results at age 8 years of early intervention for low-birthweight. JAMA 1997;277:126-32. 26. Wolke D. Environmental neonatology (annotation). Arch Dis Child 1987;62:987-8. 27. Wolke D. The preterm responses to the environment—long-term effects? In: Cockburn F, editor. Advances in perinatal medicine. Carnforth: The Partheon Publishing Group; 1997. p. 305-14. 28. Jacobsen T, Gronvall J, Petersen S, Andersen GE. “Minitouch” treatment of very low-birth-weight infants. Acta Paediatr 1993;82:934-8. 29. Tyson JE, Younes N, Verter J, Wright LL. Viability, morbidity, and resource use among newborns of 501- to 800-g birth weight. National Institute of Child Health and Human Development Neonatal Research Network. JAMA 1996; 276:1645-51. 30. Wolke D. Psychological development of prematurely born children. Arch Dis Child 1998;78:567-70. 31. Als H, Lawhon G, Brown E, Gibes R, Duffy FH, McAnulty G, et al. Individualized behavioral and environmental care for the very low birth weight preterm infant at high risk for bronchopulmonary dysplasia: neonatal intensive care unit and developmental outcome. Pediatrics 1986;78:1123-32. 32. Als H, Lawhon G, Duffy FH, McAnulty GB, Gibes-Grossman R, Blickman JG. Individualized developmental care for the very low-birthweight preterm infant—medical and neurofunctional effects. JAMA 1994; 272:853-8.

533