The Role of Neonatal Neuroimaging in Predicting Neurodevelopmental Outcomes of Preterm Neonates

Brain Injury among Preterm Infants Diagnosed by CUS

A grading approach to categorize severity of intracranial hemorrhage (ICH) in preterm infants was published in the late 1970s. In this system, hemorrhage confined to the germinal matrix is defined as grade I, into the ventricle without dilation as grade II, into the ventricle with dilation as grade III, and with parenchymal hemorrhage as grade IV, which in the original description was defined as accompanying intraventricular hemorrhage (IVH). This system remains broadly applied, both for research and for family counseling. Using these traditional definitions, severe ICH may be decreasing over time among extremely preterm infants. The Eunice Kennedy Shriver NICHD Neonatal Research Network (NRN) reviewed from 1993-2012 maternal and neonatal care, as well as morbidities and mortality among infants born at 22+0/7 weeks’ to 28+6/7 weeks’ gestational age in one of the participating academic US centers. Among extremely preterm infants overall, the proportion with severe ICH (defined as grade III or IV IVH) decreased from 1993-2012 (19%-13%), with significant reductions for infants born at 26 weeks’, 27 weeks’, and 28 weeks’ gestational age (GA) but not for those 22-25 weeks’ GA. Cystic periventricular leukomalacia (cPVL) also decreased for infants 26, 27, and 28 weeks’ GA, with 2012 rates 4%, 3%, and 2% respectively.

However, the complexity of the interpretation of CUS findings, evolving recognition of neuropathologic etiologies of intraparenchymal echogenicity or periventricular hemorrhagic infarction (PVHI), and limitations of outcomes prediction with any single neuroimaging or other finding, should give clinicians cause for prudence and careful consideration in using these classifications alone for prognostic guidance. The grading system does not include consideration for nonhemorrhagic periventricular echogenicity or echolucency such as cPVL, nor does it include ventricular dilation not associated with hemorrhage, or cerebellar hemorrhage, all of which have been associated with neurodevelopmental impairment. In some studies, persistence of periventricular echodensity or “flaring” on CUS is assessed, a finding that has been suggested to be predictive of neuromotor outcome. Since the initial description of these grades, periventricular hemorrhagic infarction (PVHI) has been proposed to be caused by impairment of venous drainage leading to venous infarction, with subsequent hemorrhagic evolution in periventricular white matter, thus may exist as an isolated finding. Furthermore, the extent and severity of the injury associated with PVHI is not further quantified. Nevertheless, strong associations between major ICH and adverse neurodevelopmental outcomes among preterm infants have been described in numerous single-center, multisite, and population-based analyses. Detailed evaluation of these associations, however, may point more to a mechanism via white matter (WM) injury rather than hemorrhage into a ventricle in and of itself.

CUS Findings and Neurodevelopmental Outcomes

The focus of many investigations has been on exploring the prognostic capabilities of major CUS findings with cerebral palsy (CP). The majority of studies have found significant associations of severe CUS abnormalities with neuromotor outcomes at 18-36 months. The EPIPAGE study followed a French regional cohort of 22-32 weeks’ GA infants to 2 and 5 years of age. Cerebral palsy (CP) was determined by questionnaires sent to pediatricians. At 2 years, among children with a history of white matter abnormalities (defined as PVL, ventricular dilation, or intraparenchymal hemorrhage or cyst), 24% were diagnosed with CP; 57% of those with cystic PVL were diagnosed with CP. The presence of cerebral lesions on neonatal CUS was independently associated with CP at 5 years. At 5 years, among those with cystic PVL, 61% had CP; among those with intraparenchymal hemorrhage, 50% had CP. At both 2 and 5 years, among those with no CUS abnormalities, CP was diagnosed in just over 4%. In EPICure 1, a population-based study of infants 20-25 6/7 weeks’ GA born in the United Kingdom and Ireland from March 1995 to January 1996, severe CUS abnormalities (defined as parenchymal hemorrhage, cystic changes, or ventricular dilatation on the last CUS) were associated with CP (OR 4.95, 95% CI 2.25–10.85) and with severe motor disability (OR 7.15, 95% CI 2.73–18.74). Of importance, when children with motor disability were excluded, severe CUS abnormality was not significantly correlated with Bayley Scales of Infant Development 2nd edition (BSID-II) mental developmental index (MDI) score. The NICHD Neonatal Research Network (NRN) reported the 18-22 months’ corrected age outcomes of a cohort of extremely low birth weight (ELBW) infants born in the 1990s and also found that after adjusting for numerous confounding variables, grade III or IV IVH (OR 2.4, 95% CI 1.8–3.1) and cPVL (OR 10.5, 95% CI 7.2–15.2) were associated with moderate to severe CP. Among extremely preterm infants <25 weeks’ GA in the NRN, IVH grade III or IV, and cystic PVL were also independently associated with moderate to severe CP. The Extremely Low Gestational Age Newborn (ELGAN) study was a multicenter study designed to identify characteristics and exposures that increase the risk of neurodevelopmental impairments at early childhood follow-up among children born before 28 weeks’ gestation. Three CUSs were required at specified age ranges during NICU hospitalization. At 2 years, intraparenchymal hemorrhage was strongly independently associated with moderate to severe CP (RR 4.2, 95% CI 2.1–8.1), but white matter injury on CUS as defined by echolucency (RR 16, 95% CI 7.6–32) and ventriculomegaly (RR 11, 95% CI 5.5–21) demonstrated even stronger associations. In multivariable analyses adjusting for numerous confounders, isolated IVH without white matter injury did not appear to be associated with neurologic or developmental impairment at 2 years as defined by BSID II psychomotor developmental index (PDI) and MDI scores. Only when accompanied by white matter lesions was IVH independently associated with significantly increased risk for CP and no more than a minimal increased risk for adverse developmental outcome.

These studies and others show that severe neonatal CUS abnormalities are associated with subsequent neuromotor impairment. Nevertheless, the risk for CP is not completely eliminated with the finding of a normal CUS, although it is reassuring. In the EPIPAGE study, only 4.4% of those with normal CUS went on to be diagnosed with CP at 2 years. However, evaluating the data in a different way, of the 164 children diagnosed with CP, more than one-third had no abnormality on neonatal CUS. In the ELGAN study, almost half of the children with CP at 2 years had all normal CUSs, and the positive predictive value (PPV) of ventriculomegaly or echolucency for moderate or severe CP was poor. Nevertheless, with meticulous technical attention and serial CUS imaging, impressive predictive capability for CP has been reported. In a single center, de Vries et al. reported 76% sensitivity and 95% specificity of CUS abnormalities for CP at 2 years for patients <32 weeks’ GA. Of importance, among those with major CUS abnormalities who developed CP, approximately 30% were noted only after 28 days of age on serial imaging, including some cystic changes that were found only on late scans or were observed to collapse or coalesce over time. In longer-term follow-up at 5 years of age in the EPIPAGE cohort, major CUS abnormalities remained strongly associated with motor challenges, with CP seen in 61% of those with cPVL and 50% in those with intraparenchymal hemorrhage, compared with 4% in infants with normal neonatal CUS.

Although severe CUS findings are associated with subsequent adverse motor findings, CUS findings alone are poorly predictive of early developmental outcomes or later childhood cognitive and learning outcomes unless more detailed CUS and other factors are taken into account. Hack et al. found that severely abnormal CUS findings among ELBW infants was independently associated with abnormal neurologic outcome but not with BSID II MDI <70, at 20 months. However, about 40% of ELBW infants with neurologic abnormality at 20 months did not have abnormal CUS, and about half of those with abnormal CUS did not go on to have neurologic abnormality. Similarly, among ELBW infants in the NICHD NRN with at least two neonatal CUSs during hospitalization and all CUSs reported as normal, either BSID II MDI <70 or CP was still present in 29% at 18-22 months. Further analyses demonstrate that the presence of severe IVH on CUS accounts for a very small fraction of the variation in major handicap or low MDI score, and models that include clinical variables predict neurodevelopmental outcomes significantly better than those with CUS variable alone. Even with serial CUS and evaluation of detailed findings, major CUS abnormalities have not been found to be strongly associated with cognitive delay at 2 years. In the 8-year follow-up of the EPIPAGE cohort, ~40% of those with major neonatal CUS abnormalities had no significant cognitive or learning challenges identified, whereas 30-40% of those with no neonatal CUS abnormalities had moderate to severe challenges. This underscores the need for long-term surveillance through childhood for all born extremely preterm regardless of neuroimaging findings.

Potential Limitations to Interpretation

Details pertaining to severity of IVH and PVHI may be important to understanding implications for outcome, as can the presence or absence of other clinical findings and morbidities. Any grade III and IV ICH have often been combined into a single “severe” category both for the purposes of research and clinical counseling. However, Bassan and colleagues scored the severity of PVHI on CUS on the basis of extent, bilaterality, and midline shift. Two-thirds of those with PVHI had significant neurodevelopmental impairments, but severity of outcome was correlated with CUS PVHI severity score. Merhar et al. found that the rate of neurodevelopmental impairment increased with unilateral to bilateral ICH, as well as with sepsis and postnatal steroid exposure. Of note, the group also found that although infants with bilateral grade IV had substantially worse BSID-II MDI and PDI scores than those with unilateral, those with bilateral and unilateral grades I–III IVH had similar outcomes, and those with grade III had similar outcomes to those with grades I-II. Among preterm infants surviving after intraparenchymal hemorrhage in a single center spanning a more recent era when Bayley Scales 3rd edition (Bayley-III) was in use, Tsai and colleagues found significant predictors of CP, including ventriculomegaly, intraventricular echodensity, and ventricular shunt, but no clinical, imaging, or sociodemographic factors were significantly associated with low cognitive scores. In a retrospective analysis of extremely preterm infants diagnosed with grade III or IV ICH in the NICHD NRN with 18-22 month follow-up including Bayley-III, bilaterality of the findings as well as the presence of PHVI were associated with death or impairment, although birth weight, presence of shunt, and exposure to postnatal steroids influenced prediction. Conversely, “low-grade” (grades I and II) hemorrhage is often considered to confer no additional risk for adverse neuromotor or developmental outcomes. In a single-center study, ELBW infants with uncomplicated grade I or II IVH were reported to have poorer neurodevelopmental outcomes at 20 months than those with normal CUS, even after adjusting for confounding risk factors. However, a multicenter study of infants <27 weeks’ GA born 2006–2008 found that 18-22 month outcomes did not differ for those with grades I and II than for those without hemorrhage. A large, single-center, case-control study of infants 24-32 weeks’ GA found no difference in neurodevelopmental outcomes between those with grade I-II and those with normal CUS, whereas grade I-II was found to be associated with adverse neurodevelopmental outcomes among 23-27 weeks’ GA infants in a regional cohort in Australia. In evaluating associations of neonatal CUS findings with later outcomes, the National Brain Hemorrhage Study cohort isolated germinal matrix, and IVH without ventricular enlargement was not associated with adverse outcomes at 6 and 9 years. In even longer-term follow-up, utilizing data from the low birth weight sample of the Infant Health and Development Program study, grades I and II were also not independently associated with cognitive, behavioral, or academic outcomes at 3, 8, or 18 years of age. These disparate findings with regard to low-grade hemorrhage may be reflective of differences in patient population, era, brain imaging and outcomes, ascertainment tools, and details of brain injury.

Interpretation of the applicability of studies relating severe or low-grade CUS findings with neurodevelopmental outcomes may also be challenged by the validity of findings. The NICHD NRN assessed interobserver reliability of CUS findings between two central readers and accuracy of local compared with central readers. Agreement between central readers was high for major CUS findings such as grade III or IV IVH and degree of ventriculomegaly (kappa = 0.84 and 0.75, respectively) but much worse for lower-grade IVH (kappa = 0.4). The sensitivity of local reader interpretation was also excellent for severe IVH (88%-92%) but poor for grade I or II IVH (48%-68%). Also, in many studies, no specific CUS protocols were required, and the frequency of CUS was not specified. A primary guideline for CUS screening in the United States is the Practice Parameter for Neuroimaging of the Neonate in 2002, which recommends screening with CUS for all infants with GA <30 weeks at 7-14 days, and “optimally ” again at 36-40 weeks’ postmenstrual age. However, more frequent and detailed surveillance protocols using high-resolution techniques may increase the sensitivity of CUS for identifying infants at high risk for adverse outcomes.

It is clear that CUS alone can be a valuable tool to assist in prediction of neuromotor outcomes. But studies of neonatal CUS and later outcomes have implicated white matter injury as the important underlying etiology linking abnormal brain imaging findings with adverse neurodevelopmental outcomes among preterm infants. This has led to the suggestion that white matter injury can be better recognized and more comprehensively characterized; we may be better able to predict adverse motor and developmental outcomes and potentially anticipate specific needs for early evaluation and intervention. More importantly, a better understanding of the connection between perinatal events and brain injury may allow for prevention or amelioration of brain injury in the preterm. These broad hypotheses have formed the basis for investigation of magnetic resonance imaging (MRI) as a routine or adjuvant neuroimaging modality for preterm infants.