Changing Definitions of Long-Term Follow-up: Should “Long-Term” be Even Longer?

Abstract

There have been amazing changes in outcomes of preterm (PT) infants in the past 200 years.¹ Whereas early studies reported only survival rates, Dr. Hess published the first outcome study of PT infants in Chicago in 1953.¹ Dr. Lubchenco then published the 10 year follow-up of premature infants born in 1947 to 1953 and identified a 68% handicap rate.² As a result of these early studies, the importance of evaluating NICU graduates both for surveillance and as an outcome of trials was recognized. During the 1970s there was a gradual expansion in the number of follow-up programs in the United States (US) with an increasing number of follow-up studies published. In the 1980s the importance of multicenter clinical research networks was recognized and the NICHD Neonatal Research Network (NRN) was initiated in 1986. Follow-up protocols, definitions, and outcomes have evolved over the last 30 years and will be reviewed with a focus on NICHD NRN studies.

Eunice Kennedy Shriver NICHD Neonatal Research Network Follow-Up Study Group

The NICHD NRN was initiated as a multicenter effort in the US with the main objective of providing a registry of uniformly collected baseline and morbidity and mortality data information to provide the basis for planning and implementing clinical trials. The NICHD NRN Follow-up Study Group was added in 1993 to provide neurodevelopmental follow-up for trials, and for those meeting NRN Follow-Up Study criteria, which now includes infants <27 weeks’ gestation born at a NRN site, included in the NRN general data collection, and with appropriate site-specific institutional review board approval and consents. The NRN research objective is to utilize standardized protocols and definitions in order to provide outcomes for randomized control trials, longitudinal outcome studies, ongoing surveillance and observational studies. The NICHD NRN’s primary focus is to improve the outcomes of premature infants and full term high risk infants.

Examiners and Certification Procedures in the NICHD NRN

An important component of the NRN follow-up protocols is the certification of NRN examiners in assessment procedures. Training of multiple examiners to code neurodevelopmental outcomes in young children is challenging because of heterogeneity of neurologic findings ranging from mild to severe which may not clearly fall into a specific category.³ These challenges have been addressed with a rigorous training and certification protocol. Trained examiners are certified annually for the Bayley Scales of Infant Development III (BSID-III) and the neurologic exam. Each center designates a primary BSID-III examiner and a primary neurologic examiner, who are responsible for training and certifying additional examiners at their center. Each year, the primary BSID-III examiner from each center submits a video of him/her performing the exam with a 22–26 month old child to one of four NRN BSID-III Gold Standard examiners who provide feedback on administration and scoring. The annual certification process for the BSID has remained unchanged.

The annual certification process for the neurologic exam has evolved over time to improve quality and efficiency, identify problematic areas for training, and enhance inter-rater agreement as described in detail elsewhere.³ In the early years of the NRN (1994–1999), patients with a variety of neurologic diagnoses were examined during centralized training. At present, the center primary neurologic examiner submits a video each year of him/her performing the exam with a child at 22–26 months corrected age (CA). The videos are reviewed by two NRN neurologic exam Gold Standards who choose 5–6 high quality exams with a spectrum of neurologic findings to place on the annual certification DVD. Prior to the annual training workshop, primary examiners view and score the exams on the certification DVD and their responses are keyed and transmitted to the RTI International data coordinating center for inter-rater agreement analysis. The annual training workshop provides targeted training on problematic items and concludes with examiners scoring a “test” video. Examiners receive feedback on the video they submit, their scoring of the certification DVD, the “test” video. This process has rendered improvements in inter-rater agreement as well as pre- and post-training within a single year.³

NICHD Follow-up Study Cohorts and Changes in NRN protocols; 1993 and 2016

Population

The NRN maintains a generic database (GDB) that includes all infants on whom neonatal data are collected. Infants in the GDB (401–1500 grams) whose follow-up window opened before January 1, 2008 were eligible for follow-up at 18–22 months CA if they had a birth weight of 401–1000 grams. Infants whose follow-up window opened January 1, 2008 or later are eligible for follow-up if they fit either of the following criteria: GDB inborn infants born at ≤ 27 completed weeks gestational age (up to and including 26 6/7 weeks), or larger PT infants enrolled in a randomized trial or observational study with follow-up as a predefined primary or secondary outcome. This change was initiated in response to the improved survival of infants at the limits of viability.^4–6

Age

For GDB infants born before July 1, 2012, the follow-up assessment was at 18–22 months CA. For infants born July 1, 2012 or later, the follow-up assessment is completed at 22–26 months CA which is consistent with other national and international networks.

Neurodevelopmental assessments and definitions

Neurodevelopmental outcomes have become increasingly important as survival of even the most extremely preterm (EPT) and highest risk term infants is increasingly in prospective studies of EPT infants have often taken place during the toddler period or in very early childhood (18 months of age corrected for prematurity to 3 years). At these ages, a battery of exams and tests can be performed including neurosensory and neurologic exams to assess for blindness, deafness and cerebral palsy (CP), and standardized cognitive and psychomotor assessments such as the Bayley Scales of Infant Development (BSID).⁷ Interpretation and comparison of studies, however, may be difficult. Many studies report outcomes by categories, and frequently combine any adverse finding from various aspects of the overall visit to attain an “impaired” or “not impaired” status. The definitions of “adverse” outcomes may not be consistent across studies and cohorts; indeed, the definitions of the individual components of “impairment” such as CP, blindness, deafness, and developmental delay, may also differ across studies. The developmental tests themselves may change substantially over time and with new editions (i.e., BSID-II to BSID-III), from making comparison of outcomes to an earlier era complex.

Development and NDI

The NRN definition of neurodevelopmental impairment (NDI) has evolved over time and has included a composite of neurologic, developmental, vision, and hearing status. Changes were necessitated in response to the implementation of the BSID-III. Not all NRN studies use the composite outcome and in certain protocols it is more appropriate to identify specific language, cognitive, or neurosensory impairments as the primary outcome. Current component definitions include the following. Moderate to severe CP includes children with tone/reflexes abnormalities and a Gross Motor Function Classification Score ≥2.⁸ Vision impairment is defined as bilateral acuity < 20–200 with amplification, and hearing impairment is defined as bilateral permanent hearing loss that does not permit the child to understand the examiner directions ± amplification.

Development

Most follow-up studies of extremely low birth weight (ELBW) infants prior to 2008 used the (BSID II).⁹ The BSID -II scales were previously the most commonly reported assessment of development from 4 to 36 months of age and provides information for cognitive [the mental developmental index (MDI)) and motor (the psychomotor developmental index (PDI)] domains. Bayley scores of 100 ± 15 represent the mean ± 1 standard deviation (SD) of a population of normal infants born at term. A score <70 (2 SDs below the mean) is interpreted as evidence of developmental delay. The BSID-III is currently used for assessments from 1 to 42 months of age.⁷ The BSID-III was developed, in part, to separate cognitive from language domains to eliminate the bias imposed for children residing in a bilingual or non-English speaking household. In contrast to the BSID-II, the BSID-III consists of three domains: the Cognitive, Language and Motor Composite scores, in addition to subscores for receptive and expressive communication, fine and gross motor skills. A limitation of the BSID-III is that cognitive and motor scores have been shown by many investigators to be higher than BSID-II scores.¹⁰ The reasons for this difference are not entirely clear. Although the BSID-II has not been a stellar predictor of school age outcomes, the predictive value of the BSID-III remains unknown. Higher mean values of the BSID-II have necessitated modifications of the cut-points of the BSID-III composite scores. Current NICHD NRN categories are moderate delay 70–84, severe delay < 70 and profound delay ≤ 54.

Behavior

The first standardized test for behavior implemented by the NRN was the Brief Infant Toddler Social Emotional Assessment (BITSEA) ¹¹ which assesses emotional difficulties for infants 12 to 36 months and was appropriate for the NRN assessment age of 18–22 months. Parents rate 49 problem scale items and 11 competence scale items and two standardized scores are derived for the problem and competence scales. With the change in the 22 to 26 month assessment age in 2008, the NRN investigators changed to the Child Behavior Checklist (CBCL) ¹² to better classify a spectrum of behavior disorders. The CBCL for ages 1.5–5 years is administered to parents and includes 99 items that describe specific kinds of behavioral, emotional, and social problems that characterize preschool children. In response to the concerns regarding increasing rates of autism spectrum disorder (ASD) in the US, the Pervasive Developmental Disorders Screening Test, Second edition (PDDST-II) Stage 2¹³, a screening tool for children 12–48 months of age was added in November 2008. The Stage 2 parent report measure was used because it is specifically designed for use in developmental clinics. NRN investigators have examined behavior outcomes using these tools.^14–16 The current protocol does not include the PDDST-II.

CHALLENGES TO EARLY NEURODEVELOPMENTAL OUTCOME ASSESSMENT

Very early childhood developmental and neurologic outcomes evaluations should only be considered as a first step to comprehensive follow-up; assessments into school age, adolescence, and adulthood are critically important to understand the functional and societal outcomes of former PT infants.^17–23 Concerns regarding changes in cognitive and academic challenges over time, and an increasing recognition that physical and environmental effects as well as early intervention approaches may modify recovery, underscore the need for later assessments.^24–26 Some neurocognitive, executive function and behavioral challenges may only be detected at school age; even recognizing that such learning and attention problems may occur in PT infants is a critical step to ensuring adequate support and services for families and teachers to help children achieve their best possible outcomes. Evaluation of neuromotor outcomes throughout childhood is critical, as later neuromotor and coordination problems such as developmental coordination disorder (DCD) are prevalent among school-age children born EPT compared with term, and can be associated with other functional challenges and academic difficulties.^{27, 28}

Despite the many challenges, changes and provisos, a substantial body of literature exists on early neurologic, cognitive, neurodevelopmental and functional outcomes of PT infants published over the past decades ranging from small or single-center cohorts to multi-site observational analyses to follow-up after neonatal intervention. We do not attempt to present a comprehensive review of all published neurodevelopmental outcomes studies in the following sections of this chapter, but rather we will concentrate on results from the NICHD NRN, and also highlighting recent large cohorts and regional population-based studies (TABLE 1).

TABLE 1.

Early neurodevelopmental outcomes: Selected recent extremely preterm cohorts

	NICHD NRN31	NICHD NRN 5	VICS 2005145	Japan NRN147	EPICure 2 148	EXPRESS38
Study group	<25 wk EGA	22–26 wk EGA	22–27+6/7 wk EGA	22–25 wk EGA	22–26+6/7 wk EGA	<27 wk EGA
Birth years	2002–2004	2006–2011	2005	2003–2005	2006	2004–2007
Age corrected for prematurity	18–22 months	18–22 months	2 years	36–42 months (chronological)	3 years	2.5 years
# (% follow-up of eligible survivors)	405 (93%)	2630 (92%)	163 (95%)	562 (72%)	576 (55%)†	415 (90%)
Blind	2.2%	0.4%	0	4.6%‡	1%	0.9%
Deaf/require aids	4.3%	1.4%	2.5%	Requires aids: 1.7%	Aids improve 0.2% do not Improve: 5%	Aids improve: 0.2% do not Improve: 0.7%
Developmental impairment	BSID-II MDI<70: 51% MDI<50: 19%	BSID-III Cognitive 70–84: 16.5% <70: 9.3%	*DQ None/1SD: 52% 1–2SD: 32% 2–3SD: 12% >3SD: 4%	*KSPD DQ <70: 35% <50: 11%	*Predicted MDI 70–84: 34% <70: 30%	*Cognitive/language None: 55% Mild: 25% Moderate: 11% Severe: 9%
CP or motor delay	Moderate: 8.7% Severe: 6.2%	Moderate: 3.4% Severe: 2.5%	Any CP: 9.8%	Any CP: 13.7% Profound CP: 8.2%	Any CP: 14% Moderate motor:3% Severe motor: 5%	Mild: 2.9% Moderate: 2.9% Severe: 1.3%
Overall disability or impairment	Any NDI: 58.5% Profound NDI:17.5% None/ minimal: 21.9%	Moderate NDI: 23.5% Severe NDI: 13.6%	None: 51% Mild: 29% Moderate: 16% Severe: 4%	§Any: 42.4% §Profound: 22.6%	None/mild: 75% Moderate: 12% Severe: 13%	None: 42% Mild: 31% Moderate: 16% Severe: 11%

^*For VICS: DQ = developmental quotient compared with a contemporaneous NBW control group; for EPICure 2: Predicted MDI (Mental Developmental Index) BSID 2^nd Edition from BSID-III; for EXPRESS: aggregated BSID-III cognitive and language score information, with mean and SD relative to a contemporaneous 37–41 week GA control group; for Japan NRN: formal evaluation by the Kyoto Scale of Psychological Development (KSPD) was available for only 318.

^†Face to face study examiner evaluations occurred in 55.3%. Multiple imputation from complete perinatal, neonatal, and sociodemographic information estimated outcomes for the entire cohort.

^‡Defined as no functional vision in one or both eyes

^§Includes 173 with informal assessments of developmental delay by pediatricians, without formal developmental testing

NEURODEVELOPMENTAL OUTCOMES OF EPT, ELBW INFANTS IN THE NICHD NRN

The NICHD NRN reported outcomes at 18–22 months of age corrected for prematurity for ELBW cohorts born 1993–94, 1995–96, and 1997–1998, stratified by EGA.²⁹ In regression analyses, rates of moderate-severe CP did not change over time during these eras, but not surprisingly, rates were significantly different between the gestational age groups. However, for developmental outcomes (assessed in these eras by BSID-II MDI and PDI) and the composite outcomes of NDI, a significant decrease in rates of MDI<70 and NDI was noted over the three eras, primarily explained by differences between rates in the earliest (1993–94) compared with the second (1995–96) eras. Although the rates of CP and NDI reported by the NICHD NRN and by other groups during these eras may be concerning, gross motor function in many of these children may in fact be relatively normal. In separate analyses of Gross Motor Function Classification System (GMFCS) findings in ELBW survivors born 1995–98, Vohr, et al. found many children with CP had normal or mildly delayed gross motor function.³⁰

More recently, the NICHD NRN reported 18–22 months CA outcomes of infants inborn at an NRN site at <25 weeks’ gestation during two epochs, 1999–2001 (epoch 1), and 2002–2004 (epoch 2).³¹ Profound NDI in this study was defined as MDI<50, or GMFCS level 4 or 5, while “unimpaired or minimally impaired” was defined as having none of the following: moderate-severe CP, bilateral severe hearing loss or blindness, MDI or PDI <85. Only about 25% of both epoch follow up groups were composed of children born at or less than 23 weeks’ gestation. Although results showed an apparent absolute increase in NDI of 50% to 59% from epoch 1 to 2, epoch was not found to be associated with NDI on multivariable analyses adjusting for baseline variable differences between the groups. Profound NDI was not increased between epoch 1 (17%) and epoch 2 (18%). Rates of adverse outcomes were higher for children <23 weeks’ than 24 weeks’ gestation for both epochs, although patient numbers were small for rarer outcomes. Despite more aggressive perinatal management with increased cesarean section delivery (41% to 49%), and dramatic reduction in postnatal steroid use (64% to 33%) from epoch 1 to 2, no significant improvement in neurodevelopmental outcomes were observed. Nevertheless, approximately 22% of survivors in both epochs were unimpaired or minimally impaired at 18–22 months CA, which is slightly higher than the 16% unimpaired survival rate among ELBW children previously reported, although the definition of unimpaired was more stringent in that analysis.³² This study differed from other international cohorts in that it narrowly focused on infants of less than 25+0/7 weeks, and they were evaluated at an earlier age. Of note, since this study, the NICHD NRN has changed the window for follow up to 22–26 months CA. Furthermore, these cohorts are not population-based, but rather representative of several US academic centers. As with previous studies and analyses from the NICHD NRN,³³ center differences in outcomes were also observed, suggesting that practice variability beyond what can be explained by perinatal risk factors and neonatal morbidities may play an important role.

Between-hospital variation in outcomes was further explored by Rysavy, et al. among infants inborn at an NRN site at <27 weeks’ gestation between 2006–2011, assessing survival and NDI at 18–22 months.⁵ During this period, the BSID-III was utilized. Impairment was defined similarly as above, with the exception that BSID-III cognitive scores of <70 were included in the severe NDI, and scores of 70–84 were included in the moderate NDI definitions. Outcomes were evaluated in relation to hospital rates of “active treatment” defined as potentially life-saving treatments initiated after delivery. For all infants, overall rates of survival without severe impairment were (mean) 3.4%, 17.9%, 44.7%, 61.1%, and 75.6% for 22, 23, 24, and 25 weeks’ gestation. However, there significant differences among hospital rates of active treatment for those born at 22, 23, and 24 weeks, which accounted for a substantial proportion of variation in outcomes for children born at that gestational age range. The BSID-III was utilized during this study period, and there was no term control group. As noted previously, using BSID-III test cut points have been reported to underestimate impairment in EPT infants^{10, 34} and results are difficult to compare to those of previous instruments.

FACTORS ASSOCIATED WITH NEURODEVELOPMENTAL OUTCOMES

Early childhood neurologic and developmental disability in PT infants may be associated with an enormous array of biologic, environmental, and iatrogenic factors. Furthermore, the risk factors themselves may be tightly associated with other variables, and one factor may modify or potentiate the effect of others; therefore, analyses can be complex and challenging to interpret. Some studies from the NICHD NRN have suggested that little of the variance in outcomes can be explained by identified major risk factors,³⁵ and that clinical factor models predict neurodevelopmental outcomes better than cranial ultrasound (CUS) factor. Nevertheless, several variables have been consistently recognized to be associated with adverse neuromotor outcomes and overall early childhood impairment among ELBW and very preterm (VPT) infants. Risk factors specifically for developmental or cognitive delay present greater challenges, in part due to differences among studies with respect to definition of delay, and to social-environmental interactions.

Antenatal, demographic, and social risk factors

Events and interventions prior to delivery may have significant influences on EPT neurodevelopmental outcomes. Antenatal steroid treatment has been shown to improve lung maturity and reduces neonatal morbidities including respiratory distress syndrome, necrotizing enterocolitis, severe ICH, and death.³⁶ Antenatal steroid exposure also appears to be associated with reduction in NDI at 18–22 months,²⁹ and death or NDI, among those born at 23 to 25 weeks.³⁷ Location of delivery may also be associated with death or NDI. As noted previously, NICHD NRN studies have demonstrated center differences with respect to neonatal intervention, death or NDI, and NDI among survivors.^{5, 33} Serenius, et al. and the EXPRESS group,^{38, 39} demonstrated significant regional variation in obstetric and neonatal intervention approaches in Sweden; in regions with more aggressive perinatal intervention, the risk of death or NDI among 22–24 weeks was reduced, without increase in NDI among survivors. A distinct male disadvantage for death, short-term morbidities, and neurodevelopmental outcomes including cognitive delay, cerebral palsy, and motor impairment has been frequently reported for extremely PT infants.^{29, 40, 41} The reasons for this gender-specific vulnerability are obscure, but measurable risk factors and events do not appear to explain differential risk for adverse neurodevelopmental outcomes for boys.⁴¹ Social disadvantage and lower level of maternal education is associated with increased rates of adverse early neurodevelopmental outcomes.^{16, 31, 42–44} Maternal race and ethnicity also have an impact on early EPT language outcomes, even after adjustment for differences in neonatal morbidities and maternal education.⁴⁵ NICHD NRN investigators have also shown that EPT children whose primary language is Spanish have similar cognitive but lower language score than those whose primary language is English, thus suggesting that English language based testing tools may introduce bias.¹⁴

Infection, necrotizing enterocolitis, growth and nutrition

Extremely PT infants are at high risk for late-onset infection, although rates infection has decreased in the NICHD NRN between 2005–2012 for EPT infants across gestational age weeks.⁶ Both neonatal sepsis and severe necrotizing enterocolitis (NEC) have been shown to be associated with adverse motor, cognitive, and growth outcomes at 18–22 months,^46–48 with high risk for death or NDI among those with Candida sepsis or meningitis.⁴⁹ White matter injury is a key component in the path from infection or NEC to poor neurodevelopmental outcome. The inherent vulnerability of the pre-oligodendroglial cell places the PT infant at high risk.^{29, 30} Factors known to be injurious or directly toxic to developing white matter, such as systemic and cerebral hypoperfusion, ischemia-reperfusion, and cytokine elucidation during a systemic inflammatory response, can be expected during the clinical course of infection or NEC. Neonatal sepsis and NEC have been linked with progressive or higher rates of WMI on MRI; adverse 2-year outcomes associated with these morbidities may be mediated by WMI.^{50, 51}

Numerous clinical investigations among EPT infants have demonstrated that better nutritional support is associated with reduced rates of extrauterine growth retardation (EUGR), and that improved growth and better nutritional support during the NICU hospitalization is associated with improved neurodevelopmental outcomes.⁵² In a prospective observational study of ELBW infants in the NICHD NRN, those with higher in-hospital growth velocity rates had decreased clinical morbidities and neurodevelopmental impairment, and were significantly less likely to fall below the 10^th percentile for growth at 18–22 months CA.⁵³ In regression analyses adjusted for demographic and clinical confounders, NICU growth velocity remained significantly associated with growth and neurodevelopmental outcomes at 18–22 months CA. Additional analyses from the NICHD NRN Glutamine Trial⁵⁴ found that the influence of critical illness on the risk of adverse outcomes was mediated by total daily energy intake during the first week of life.⁵³ Vohr, et al. found a significant, independent association of breast milk provision during NICU hospitalization with 18–22 month CA outcomes, demonstrating that for every 10-ml/kg per day increase in breast milk ingestion, MDI increased by 0.53 points, and rate of re-hospitalization decreased by 6%,⁵⁵ and the positive effect of breast milk on outcome persisted to 30 months CA.⁵⁶ These studies and others have led to evidence-based standardized feeding guidelines, including early parenteral and enteral nutrition, strong support of breast milk provision, and focus on postnatal growth targets. Implementation of such guidelines have resulted in improvement of nutritional milestones, decreased rates of severe EUGR, decreased days of parenteral nutrition, and decreased rates of NEC.⁵⁷

Bronchopulmonary dysplasia (BPD)

Extremely PT infants are at risk for BPD, which is often defined as requiring oxygen at 36 weeks’ postconceptional age; it is diagnosed in 30–50% of EPT infants, with rates increasing between 2009–2012 among those 26 and 27 weeks’ gestation.⁶ Those with BPD been shown to be at higher risk for all components of early neurodevelopmental impairment, and later childhood global neurocognitive impairment.^{58, 59} Perinatal infection and placental inflammation is associated with increased risk for BPD, and infants with BPD are also more likely to have other neonatal morbidities including late-onset sepsis and ICH. A cascade of events including recurrent hypoxic ⁶⁰ and hypoperfusion episodes, purportedly leading to injury of the vulnerable developing brain. In addition, postnatal dexamethasone, used in an attempt to treat BPD, has been shown to be associated with adverse neurodevelopmental outcomes particularly when initiated early, possibly directly by inhibition of brain growth.^{61, 62} Preterm infants with BPD also have respiratory challenges after NICU discharge, and impaired pulmonary function and exercise capacity through childhood. Therefore, concerted efforts to reduce BPD in the NICU could have lifelong positive effects for children born EPT. In a NICHD NRN cluster-randomized trial of benchmarking and quality improvement approaches, practices were changed at intervention sites, but BPD rates were not improved.⁶³ Nevertheless, a predictive model of BPD was developed from the detailed data collected in this trial, which is now available as a web-based tool to provide accurate estimates of BPD based on demographic variables and respiratory support by postnatal day. (⁶⁴, https://neonatal.rti.org). The NICHD NRN also performed a multicenter 2-by–2 factorial design trial, including random assignment to intubation and surfactant treatment or to CPAP in the delivery room among infants 24 to 27+6/7 weeks with a primary outcome of death or BPD.⁶⁵ Although no significant difference was found in the primary outcome between groups, secondary outcomes including postnatal steroid use and duration of ventilation were reduced in the CPAP group. At 18–22 months CA, there were also no differences between groups in death or neurodevelopmental impairment, or components of neurodevelopmental impairment,⁶⁶ but, those in the CPAP arm had less respiratory morbidity as compared with intubation and surfactant as the initial delivery room approach.⁶⁷ Given the multidimensional nature and complex challenges to understanding BPD in the current era, the NHLBI developed the Prematurity and Respiratory Outcomes Program (PROP)⁶⁸ as a multidisciplinary, longitudinal approach, contributing additional high resolution data to results from previous studies to inform development of future interventions to improve outcomes.

Brain Injury

CUS abnormalities and early neurodevelopmental outcomes

Virtually every major study of early neurodevelopmental outcomes among PT and ELBW infants has confirmed a strong association between major CUS abnormalities and adverse neurologic and developmental outcomes. Definitions of CUS abnormalities as well as specific outcomes differ among studies, however most consider IPH, ventriculomegaly (VM), or cystic changes, regardless of laterality or extent of the findings, to be severe abnormalities. In some, persistence of periventricular echodensity or “flaring” is included.^{69, 70} The diagnosis reported is frequently based on the results from a single CUS, either the “worst” or the “final” imaging study, but some prospective cohorts^70–73 include serial imaging.

The focus of many studies has been on exploring the association of major CUS findings with cerebral palsy. The Extremely Low Gestational Age Newborn (ELGAN) Study followed infants <28 weeks’ gestation from 14 institutions across 5 states in the US during 2002–2004.^{71 73} Three study CUS were performed during hospitalization, scored by study radiologists. The investigators found strong independent associations of specific CUS findings with CP. About half the children with CUS echolucency or VM developed CP, and late occurrence of VM, bilateral echolucency, and IPH or PVL were strongly predictive of quadriparesis. Isolated intraventricular hemorrhage (IVH) was not strongly predictive of CP, and major CUS abnormalities have been shown not to be associated with developmental scores. Furthermore, O’Shea, et al. and the ELGAN study group showed that only when accompanied or followed by WM lesions was IVH associated with increased risk for motor or developmental impairment at 2 years, highlighting the limited predictive value of both early CUS findings and IVH alone.⁷⁴ However, almost half of the children in the ELGAN group with CP at 2 years had completely normal CUS, and NICHD NRN investigators⁷⁵ and others^{40, 69} have demonstrated that 30–40% of those with normal CUS have neurodevelopmental challenges at 18–30 months. Nevertheless, much can be seen beyond ICH by CUS with careful, serial imaging. In a single center, deVries, et. al. found that the sensitivity and specificity of CUS abnormalities for CP at 2 years was an impressive 76% and 95% for patients <32 weeks EGA.⁷⁰ Of importance, among those with major CUS abnormalities who developed CP, approximately 30% were noted after 28 days. Using MRI at term equivalent age to refine specific CUS findings has resulted in positive predictive value (PPV) and negative predictive value (NPV) of 96% and 69%, respectively, for CP at 2 years.⁷² Major CUS abnormalities were not strongly associated with cognitive delay at 2 years.

As with all imaging modalities, interpretation may be challenging. In an analysis of CUS central and local readings in the NICHD NRN PiNO trial, although interrater reliability and accuracy was shown to be very good to excellent for severe ICH, agreement was fair or poor for more subtle findings and periventricular leukomalacia alone.⁷⁶ Cerebellar hemorrhage is increasingly recognized to be associated with neurodevelopmental disabilities in children born PT^{77, 78}, a finding that may be missed without appropriate CUS views, and transient lesions may be missed, including echodense periventricular lesions or collapsing small cystic lesions.⁷⁰ Isolated intraparenchymal hemorrhages and of course large IVH can be seen by CUS; however, not all “severe” hemorrhages are equivalent with respect to early neurodevelopmental outcomes. Characteristics of the hemorrhage including laterality, midline shift, and extent of hemorrhage,^{79, 80} as well as the presence or absence of other adverse clinical factors,⁸¹ have been shown to impact prediction of neurodevelopmental outcomes.

Conventional MRI findings and early neurodevelopmental outcomes

Conventional MRI has been used more extensively in recent years among very PT infants, both for research and clinical indications. MRI allows for a more comprehensive and detailed picture of the brain, and better delineation of deep structure and cortical injury, as well as improved detection of white matter injury (WMI). This is common among PT infants at term, and critically important to understanding the structure-function relationship of the developing PT brain, influences on later neuromotor and cognitive outcomes, and developing future neuroprotective strategies.⁸² Advanced MR imaging in PT infants at near-term have shown that subtle WM injury is associated with reduced total brain and gray matter volumes, reduced cerebellar volume, and reduced basal ganglia and thalamic volume,^83–87 which in turn, are associated with developmental impairment in childhood among PT infants.⁸⁸ These findings and others have provided evidence that WMI in the PT is associated with brain maturational disturbances, suggesting an overall link to impaired neural connectivity.⁸⁹ Thus, clinical investigations have focused on whether MRI may provide enhanced prognostic information.

Early studies attempting to compare term equivalent MRI with CUS predictive capabilities were primarily small, single center studies, and timing and approach to CUS differed.^90–92 Since that time, WMI scoring approaches have been developed, and larger cohort studies have been published, among the first of which was a multi-center effort in Australia and New Zealand comparing serial CUS with near-term MRI findings and their association with 2-year in 167 infants <30 weeks EGA.⁹³ This study demonstrated that moderate-severe WM abnormalities on near-term MRI were significantly associated with neuromotor delay and cerebral palsy, independent of CUS findings and other risk factors. Increasing WMI severity was also linearly related to worsening BSID-II MDI scores, but an independent association of moderate-severe WMI with severe cognitive delay was not reached. However, CUS was assessed only with regard to early findings including grade of ICH and periventricular cystic changes, and a substantial proportion of infants with moderate to severe WMI by MRI did not have adverse 2-year outcomes. The NICHD Neuroimaging and Neurodevelopmental Outcomes (NEURO) study was a prospective study of early and late CUS and near term MRI including approximately 500 infants <28 weeks’ gestation in NRN centers across the US, with outcomes including BSID-III assessed at 18–22 months.⁹⁴ In multivariable models, both late CUS findings reflective of WM injury and MRI findings of significant cerebellar injury remained independently associated with adverse neurodevelopmental outcomes. In models that did not include late CUS, MRI findings of both moderate to severe WMI and significant cerebellar lesions were independently associated with adverse outcomes. Early CUS findings were not associated with adverse outcomes when any late neuroimaging was taken into account. These results demonstrate the need to understand the evolution of brain injury over time in outcomes prediction rather than to rely on early findings. The NICHD NEURO cohort is being followed to 6–7 years to examine absolute and relative value of neonatal CUS and MRI in predicting neurologic, coordination, cognitive and behaviors outcomes at school age.

Controversies and questions remain as to which neuroimaging studies to perform, when to perform them, and relative values in prognosis. These are not simple questions, as the “value” of additional information may be very different in varying clinical circumstances, and for individual parents and physicians.⁹⁵ Further investigation with longer-term follow up to school age, and with advanced MR techniques including DTI, fcMRI, surface morphometry, and volumetric methods hold enormous promise to help to explore these possibilities.⁹⁶

Future studies Moderate and Late PT

Moderate PTs born at 32 to 33 weeks’ gestation and late PTs born at 34 to 36 weeks’ gestation make up the largest subgroup of PT infants contributing more than 80% of PT births in the US. There is a growing body of evidence that these infants are at increased risk of both neonatal and post-discharge morbidities. In the US in 2013, 447,875 (11.39%) of the 3,932,181 live births were PT (< 37 weeks) and 55, 443 (1%) were very low birth weight (VLBW) (< 1500 grams).⁹⁷ The second half of gestation is a critical period of brain maturation including neurogenesis, synaptogenesis, and dendritic arborization⁹⁸ and there is increasing evidence that both MPT and LPT infants are at increased risk of neurologic impairments, developmental disabilities, school failure, autism, and behavior and psychiatric problems that extend to adolescence and young adult age.^99–109

Executive Function

Executive function encompasses high-level mental processes necessary to regulate behavior and cognition for goal-directed actions, and play an important role in school achievement.¹¹⁰ They include working memory, inhibition, planning and organization, verbal fluency, and cognitive flexibility. These abilities depend on the integrity of the neural network that connects the prefrontal cortex to the brainstem, the cerebral lobes, and the limbic and subcortical regions.^111–114 Although the majority of studies of deficits in these domains have been reported in adolescents and young adults born PT,^115–119 executive dysfunctions can be assessed at 5–6 years of age.^{119, 120} Neuroimaging studies have shown decreased white matter volumes and impaired white matter gains compared with FT controls, which indicates increased vulnerability of PTs. In the Indomethacin trial¹¹⁵ adolescent PTs exhibited more problems than FT controls in executive functions, even after excluding those with neurosensory disabilities. Nosarti et al¹¹⁷ reported similar findings in a cohort of VPT young adults. Because of the important contribution of executive functions to educational achievement, further studies examining the trajectory of these functions and the role of intervention are recommended.

Mental health Outcomes

Extremely LBW survivors are at increased risk of behavior disorders including attention deficit, autism spectrum disorder, depression, and anxiety disorders.^{108, 121, 122} In a prospective, longitudinal study, VPT children were three times more likely to meet criteria for a psychiatric disorder compared to their FT counterparts at seven years of age.¹²³ A meta analysis¹²⁴ of cohorts of former PTs 10–25 years of age identified increased rates of any psychiatric illness and anxiety and depressive disorders requiring ongoing management compared to term controls. A review¹²⁵ of 15 studies of adverse behavior/psychiatric disorders of former PTs concluded there is a lack of evidence to identify specific risk factors for these disorders among VPT/VLBW survivors and concluded that additional research of large well conducted studies is needed. In view of the current high rates of mental health disorders among children and adolescents and the increased rates among former PTs, an increased focus on causation and potential interventions for these disorders is recommended.^{123, 125, 126}

Adult Medical Outcomes

Research networks in the US, currently do not have the finances to track cohorts to adult age. The expansion of electronic medical records in the US may provide the opportunity for population studies to explore adult-onset disease including obesity, pulmonary and cardiovascular morbidities, adult cognitive decline and dementia.^127–129 Epidemiologic studies from countries with population databases have identified links between prematurity and adult onset disease.¹³⁰ The study of Heinonen et al¹³¹ identified a link between late PT birth and cognitive decline at a mean age of 68.1 years. This adds one more justification to the debate of length of follow-up.

How long should high risk infants be followed?

There is an increasing body of data indicating that disability rates change between early childhood and school age.^132–134 A large population based Canadian study¹³⁴ of PT infants followed for 10 years identified that although the majority of subjects remained in their 2 year disability category, there were shifts. Early PT had a decrease in the percent with severe disability, with a small shift from moderate and severe to no or mild disability. In contrast, there was a shift of moderate and late PTs with no or mild disability at age 2 to moderate or severe disability at age 10. Although 2 year outcomes cost less, there is general agreement that school age provides a more valid outcome.^{102, 135–140}

A full-term (FT) control group

Although the NRN’s primary focus is randomized clinical trials, a full term control group has not yet been incorporated as a comparison for the generic follow-up of PT infants, primarily because of cost. As noted previously, other multicenter studies and networks have included FT controls into their models of follow-up,^{34, 115, 141–143} which has allowed for identification of impairment thresholds based on findings of a representative sample of FT children, and has demonstrated substantial limitations to relevance and validity of the findings, and may have important public policy and resource implications.^{34, 144, 145} The NRN is currently evaluating a mechanism for time-limited inclusion of a full term, normal birth weight control group going forward.

Conclusion

Although, follow-up investigations and trials have come a long way since the early studies of Hess¹ and Lubchenco,² the population of high risk infants, the complexity, types, and depth of assessments, and the duration of follow-up studies have all changed significantly. There is ample evidence for the need for long term follow-up studies.