Changes in Neurodevelopmental Outcomes From Age 2 to 10 Years for Children Born Extremely Preterm

Data were acquired from the ELGAN Study, an observational study of neurologic disorders in children born EP.¹³  All procedures for this study were approved by the institutional review boards of all participating institutions.

In the ELGAN Study, women delivered infants who had not yet reached 28 weeks’ gestational age in 14 hospitals across 5 states. A total of 1506 infants, born between 2002 and 2004, were enrolled, and 1198 survived to 10 years. We collected information about demographic, prenatal, and neonatal factors by maternal interview and review of maternal and neonatal medical records.

Trained examiners administered the Bayley Scales of Infant Development, Second Edition (BSID-II), performed a standardized neurologic examination, and assigned Gross Motor Function Classification System (GMFCS) scores at ∼2 years’ corrected age, as described previously.^14–17 

For children with an impairment that precluded BSID-II testing and for those for whom >2 test items were omitted, the Vineland Adaptive Behavior Scales Motor Skills Domain was used instead of the BSID-II Psychomotor Development Index (PDI) and the Vineland Adaptive Behavior Composite was used instead of the BSID-II Mental Development Index (MDI), as described previously.¹⁵  A total of 48 (6%) of the 802 children had missing MDI and/or PDI. Below, we refer to these surrogate measures of MDI and PDI as imputed MDI and PDI. Parents reported data about bilateral blindness and bilateral hearing impairment.

IQ was assessed with the Differential Ability Scales, Second Edition (DAS-II) at ∼10 years, as described previously.^9,18  We conducted latent profile analysis, described previously, using IQ from the DAS-II and executive function measures from the DAS-II and NEPSY-II.^19–21  Trained examiners assigned GMFCS levels (1–5). Parents responded to a survey including questions about bilateral legal blindness and need for cochlear implant(s) or bilateral hearing aids. ASD was identified by using the Social Communication Questionnaire, Autism Diagnostic Interview-Revised, and Autism Diagnostic Observation Schedule, Second Version, as described previously.²²  Children with major NDI, including profound intellectual disability, were not evaluated for ASD because diagnosis would not be valid.²³  Autism Diagnostic Observation Schedule, Second Version–calibrated severity scores, which allow for comparison of ASD severity for children with various language levels, were used to classify ASD severity levels corresponding to those defined by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition.^22,23  Level 1 ASD (“requiring support”) corresponded to calibrated severity scores of 4 to 5, level 2 (“requiring substantial support”) to scores of 6 to 7, and level 3 (“requiring very substantial support”) to scores of 8 to 10. Epilepsy was identified by using a validated seizure screen, after which there were comprehensive follow-up interviews by epileptologists, as described previously.²⁴ 

Outcomes at 2 years were classified as defined by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Neonatal Research Network to generate the Extremely Preterm Birth Outcomes Data.^3,25  We defined profound NDI as BSID-II MDI <50, PDI <50, or GMFCS 5 and moderate to severe NDI as BSID-II MDI 50 to 70, PDI 50 to 70, GMFCS 3 to 4, bilateral legal blindness, or bilateral hearing loss requiring amplification. We defined none to mild NDI as BSID-II MDI >70, PDI >70, GMFCS <3, no bilateral legal blindness, and no bilateral hearing loss requiring amplification.

An expert panel developed criteria for classifying NDI at 10 years, reaching consensus through a modified Delphi process consisting of surveys and a teleconference (Supplemental Information). The panel included 10 experts from neonatology follow-up, pediatric complex care, pediatric neurology, pediatric neuropsychology, developmental pediatrics, physical medicine and rehabilitation, and an advocate parent of children born EP. The panel members aimed to create a 10-year classification system analogous to the definition used at 2 years in the NICHD Neonatal Research Network Extremely Preterm Birth Outcomes Data using data elements available in the ELGAN Study while considering definitions of intellectual disability used by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition and American Association on Intellectual and Developmental Disabilities.^23,25,26  On the basis of recommendations from the expert panel, we classified NDI at 10 years as moderate (IQ 55–70, GMFCS 3, bilateral hearing loss requiring amplification, bilateral legal blindness, ASD level 2, or epilepsy), severe (IQ 35–54, GMFCS 4, or ASD level 3), or profound (IQ <35, GMFCS 5, or ASD level 3 combined with IQ 35–54; Supplemental Information). On the basis of the expert panel’s recommendation, we categorized children as having clinically significant epilepsy only if parents reported treatment with an antiepileptic medication at 10 years.

The unit of observation for this study was the child. We used standard summary statistics, including medians (interquartile ranges [IQRs]) and counts (percentages), to describe categorical study variables. Observations with missing data were omitted from analysis, and no imputation was performed, except as described above for BSID-II scores. Analyses used a significance level of α = .05 and were performed by using Stata SE 15.1 (Stata Corp, College Station, TX) and R statistical software (R Foundation for Statistical Computing, Vienna, Austria).

We used Cohen’s κ to test agreement of NDI classification at 2 and 10 years. We used Bhapkar’s test of marginal homogeneity to determine if the distribution of NDI differed between 2 and 10 years.²⁷  To test agreement and marginal homogeneity, the moderate and severe classifications at 10 years were combined to create ratings commensurate with those at 2 years.

A total of 1198 (80%) of individuals survived, but 132 were not assessed at the 2-year visit and 309 were not assessed at the 10-year visit, leaving 857 who were assessed at both 2 and 10 years (Fig 1, Table 1). Of these 857 children, 802 had adequate information for classification into NDI groups. Five percent were born at 23 weeks' gestation, 15% at 24 weeks, 20% at 25 weeks, 25% at 26 weeks, and 35% at 27 weeks (Supplemental Tables 5 and 6).

Children with adequate information for classification, compared with those who were not assessed or had missing data, were more frequently white, from multiple gestation pregnancies, had mothers with >16 years of education, and had mothers with private insurance (Supplemental Table 4).

Overall, 67% of children had no change in NDI classification between 2 and 10 years of age, 27% improved, and 5% worsened. Cohen’s κ was 0.34 (z = 13.2, P < .001), indicating minimal to fair agreement between the classifications at 2 and 10 years.²⁸  The distributions of NDI classifications differed between 2 and 10 years (χ²[2] = 143, P < .001).

The proportion of study subjects classified as having profound NDI decreased from 13% at 2 years to 6% at 10 years, and the proportion classified as having moderate to severe NDI decreased from 28% to 17%. The proportion classified as having none to mild NDI increased from 58% to 77% (Fig 2).

The frequency of cognitive impairment (MDI ≤70, IQ ≤70) decreased from 29% at 2 years to 16% at 10 years, composing the predominant change in NDI. The frequency of cognitive impairment at 10 years was 25% when executive function was incorporated by using latent profile analysis. The frequency of GMFCS >2 (indicating moderate to severe cerebral palsy) increased from 6% at 2 years to 12% at 10 years (Supplemental Table 12). The frequency of epilepsy was 7.6% (4% treated with medication), and the frequency of ASD was 7% at 10 years. Seven (0.9%) children determined to have epilepsy and receiving treatment at 10 years were previously classified as having none to mild NDI at 2 years. Of the 57 children with ASD at 10 years, 10 (18%) were classified as having moderate, severe, or profound NDI after a classification of none to mild NDI at 2 years. Parent-reported bilateral blindness increased from 2% at 2 years to 4% at 10 years. Parent-reported hearing loss requiring amplification increased from 2% at 2 years to 3% at 10 years (Supplemental Table 13). No children had uncorrectable hearing impairment.

We performed a sensitivity analysis using IQ <50 as a criterion for profound NDI at 10 years and IQ 50 to 70 as a criterion for moderate to severe NDI. Results were similar, except the proportion of children whose classification worsened from moderate to severe NDI at 2 years to profound at 10 years increased from 5% to 8% (Supplemental Table 7).

We performed a second sensitivity analysis using definitions of none, mild, moderate, and severe NDI at 2 and 10 years similar to definitions used in other longitudinal studies except for the addition of epilepsy and ASD (Supplemental Table 8).^5–7  Results were similar, except percent agreement decreased from 67% to 47%, and Cohen’s κ decreased from 0.34 to 0.26. The proportion children who worsened increased from 5% to 13%, and the proportion of children who improved increased from 27% to 40%.

We performed additional sensitivity analyses using a classification system of neurodevelopmental burden in children born EP.¹⁹  The predominant change in NDI was again found to be due to changes in cognitive impairment. Fewer children improved from 2 to 10 years when executive function was incorporated into the definition of cognitive impairment (Supplemental Tables 9 and 10).

The majority (70%) of children with none to mild NDI at 10 years had a stable classification from 2 years. Compared with children who improved from moderate to severe or profound NDI at 2 years, children with a stable classification of none to mild NDI had higher IQ, higher measures of executive function, less cerebral palsy, and higher quality of life scores (Table 2).

We classified 120 children as having profound NDI at 2 and/or 10 years. Most (63%) improved from classification of profound NDI at 2 years, one-third had stable classification, and a minority (10%) worsened to a classification of profound NDI at 10 years (Table 3). Of the children with a stable classification, 67% had both imputed MDI and PDI scores of 49 at 2 years compared with 15% of those who improved to none to mild.

To better inform prenatal counseling practices, we performed a subgroup analysis of 315 individuals who were born at 23 to 25 weeks' gestation with birth weights 401 to 1000 g and received either mechanical ventilation or continuous positive airway pressure at birth (representing initiation of intensive care), survived, were assessed at both 2 and 10 years, and had adequate information for classification into NDI groups (Supplemental Figure 3). Agreement was similar, but fewer children improved to a classification of none to mild NDI at 10 years after a classification of moderate to severe (53% vs 63%) or profound NDI (24% vs 36%) at 2 years compared with the primary analysis (Supplemental Figure 3).

We observed greater individual and overall improvement in NDI classifications between infancy and middle childhood than was observed in previous studies.^6,7,29,30  In sensitivity analyses, differences in NDI classification systems used in infancy and childhood only partially explained the overall higher proportion of children whose NDI classification improved between infancy and childhood and the lower proportion of children whose NDI classification worsened, as compared with previous studies.^5–7  The limited individual agreement between NDI in infancy and middle childhood confirms observations from previous studies of smaller cohorts from Sweden, the United Kingdom, and Australia; in these studies, researchers did not consider ASD and epilepsy and assessed children at younger ages.^5–7 

Cognitive impairment is the most common component of NDI in children born EP.^19,29  The BSID-II MDI has poor predictive accuracy for childhood IQ.^10,11  As previously reported, in the ELGAN cohort, the MDI at 2 years was 5 points lower than the IQ at 10 years; despite this relatively small average difference between MDI and IQ, more than one-half of the children who were classified as having cognitive impairment at 2 years (MDI <70) did not have IQ <70 at 10 years.¹⁰  In the current study, it is indicated that cognitive impairment is the largest contributor to instability in NDI over time, whereas cerebral palsy, blindness, and hearing impairment were relatively stable.^6,7,31  In previous cohort studies, researchers reported that identification of cognitive impairment is more valid if a term control group is included.^5–7  Our assessments were completed within 7 years of publication of the DAS-II, for which the reference sample was United States children, which may limit this effect.¹⁸ 

Several factors could contribute to the poor predictability of infant cognitive assessments for intellectual disability in middle childhood. Probably most important are that higher order functions cannot be comprehensively predicted or evaluated at 2 years and that the BSID-II are not intended to be a direct measure of intelligence. That children with stable classification of none to mild NDI had higher IQ at 10 years compared with children who improved to none to mild NDI (Table 2) suggests that early differences have biologically driven relevance to IQ measures at a later age. An optimistic possibility is that high-risk infants’ cognitive abilities improve as a result of developmental interventions and educational supports. Because developmentally enriched environments are associated with high socioeconomic status, results may not be generalizable to samples that differ substantially from ours with regard to the distribution of maternal education. Data on receipt of early intervention are listed in Supplemental Table 15.

Although the predictability of developmental screening in infancy is most stable in high-risk infants with the poorest performance on screening tests, predictive stability in such infants has varied widely.^5–8  Among children who, at 2 years, were classified as having profound NDI, 31% continued to have profound NDI at 10 years, whereas 13% had severe NDI, 20% had moderate NDI, and 36% had none to mild NDI. In this group, children with none to mild NDI at 10 years had a mean IQ of 88, which does not preclude the possibility of academic difficulties (Table 3). Direct comparison with previous studies is difficult because, in previous studies, researchers did not differentiate between severe and profound NDI. In EPICure and EXPRESS (Extremely Preterm Infants in Sweden Study) studies, similar results were reported, in which 60% to 64% of children had a stable classification of severe NDI and 11% to 17% improved from severe NDI to none to mild, despite many differences, such as time periods (1995 vs 2004–2007), countries (United Kingdom versus Sweden), developmental assessments (BSID-II versus Bayley Scales of Infant and Toddler Development, Third Edition), gestational age ranges (<26 vs <27 weeks), and overall rates of moderate to severe NDI (46% vs 34%).^5,6  In comparison, in the Victorian Infant Collaborative Study, it was reported that 35% of children had a stable classification of severe NDI and 38% improved to none to mild.⁷  Possible factors for higher instability in the Victorian Infant Collaborative Study include higher gestational age (<28 weeks), younger infant assessment (2 vs 2.5 years), and older childhood assessment (8 vs 6 years) compared with the EPICure and EXPRESS studies.^5–7  Our findings and those from previous studies underscore the complexity of developmental assessments in infancy, even in groups considered most stable.

Strengths of this study include a large sample selected on the basis of gestational age, high follow-up rates, NDI classification at 10 years based on consensus recommendations from an expert panel, and a comprehensive assessment of neurodevelopment by using validated assessments administered by examiners who were unaware of children’s medical history or results of previous neurodevelopmental assessments. Limitations include the lack of a term control group, the use of parent-reported outcomes for visual and hearing impairment, and the use of the BSID-II rather than the Bayley Scales of Infant and Toddler Development, Third Edition. The expert panel consensus process targeted agreement for a 10-year definition of NDI analogous to the 2-year definition used in the NICHD Neonatal Research Network Extremely Preterm Birth Outcomes Data.¹⁹  NDI definitions did not incorporate functional status or quality of life measures, which are essential components of longitudinal follow-up, especially for individuals with NDI.^32,33  The ELGAN cohort consists of infants born from 2002 to 2004, so our results may not apply to contemporary infants because of advances in survival and neurodevelopmental outcomes.

Our results have implications for clinicians who care for infants born EP and researchers who seek to improve neurodevelopmental outcomes for such infants. We confirm limited agreement between infant and childhood assessments when using a classification system familiar to many United States neonatologists.^3,4,25  When counseling parents of an infant who is expected to be born at an extremely young gestational age, clinicians should consider providing information not only on the predicted probabilities of NDI at 2 years, but also information about the proportion of children with normal range IQ later on despite poor performance on assessments in infancy. Clinicians may also consider how additional postnatal longitudinal outcomes could improve prediction.³⁴  As emphasized by Hack et al,¹¹  caution should be taken to minimize bias due to “reported high rates of cognitive impairments based on the use and presumptive validity of the BSID-II MDI.” Caution is also appropriate when interpreting findings from clinical trials involving children born EP when outcome assessments are based on neurodevelopmental evaluations before school age.

A majority of ELGAN Study infants classified as having moderate to severe or profound NDI showed improvement on assessments at 10 years compared with 2 years. For the parents at risk for delivery of an EP infant, a hopeful message can be taken from our findings that one-third of children classified as having profound NDI and nearly two-thirds of children classified as having moderate to severe NDI at 2 years had none to mild NDI at 10 years.

We thank the following team members for support of this study: editorial assistance from the NC Translational and Clinical Sciences and assistance with the consensus process from Sue Tolleson-Rinehart, PhD, and Arlene Davis, JD.

ASD

autism spectrum disorder

BSID-II

Bayley Scales of Infant Development, Second Edition

DAS-II

Differential Ability Scales, Second Edition

ELGAN

Extremely Low Gestational Age Newborn

EP

extremely preterm

EXPRESS

Extremely Preterm Infants in Sweden Study

GMFCS

Gross Motor Function Classification System

IQR

interquartile range

MDI

Mental Development Index

NDI

neurodevelopmental impairment

NICHD

Eunice Kennedy Shriver National Institute of Child Health and Human Development

PDI

Psychomotor Development Index