Neuroimaging of extremely preterm infants: perils of prediction.

Neuroimaging, primarily using cranial ultrasound (CUS), is routinely recommended in premature infants who are born at ,30 weeks’ gestational age. Most premature infants will be imaged with an early CUS examination (generally by 7–10 days of age) to assess for intraventricular or intraparenchymal hemorrhage and for evidence of early white matter injury. The role, timing, and prognostic value of later neuroimaging studies remain unclear. Recent data suggest that white matter and cerebellar injury, which may be more easily detected on MRI scans, is an important link to later neurodevelopmental impairment (NDI). This observation has led to recommendations that a routine brain MRI examination at term postmenstrual age be performed for all extremely low gestational age newborns (ELGAN) as a way to better predict the risk for NDI. In this issue of Pediatrics, Hintz et al describe the first large prospective evaluation of early (4–14 days) and near-term (35–42 weeks postmenstrual age) CUS and near-term brain MRI scans in the prediction of NDI at 18 to 22 months. In a prospective cohort of 480 infants born at ,28 weeks’ gestation who underwent all 3 scans, the primary outcome of NDI or death was assessed by certified examiners. All brain imaging studies were read centrally for evidence of intraventricular or cerebellar hemorrhage, white matter injury, and moderate to severe ventricular enlargement or the presence of a ventricular shunt. Multivariate models were constructed to include each type of scan individually and in combination, and they were assessed as predictors of NDI or death. Although the rates of abnormal scans and NDI and significant motor impairment were low, the authors found that both late CUS and MRI findings reflective of white matter injury or significant cerebellar injury were independently associated with adverse outcomes. Importantly, early CUS findings were not associated with adverse outcomes when any later neuroimaging was assessed in the model. A second important observation was that the predictive value of a combination of early and late CUS was only marginally improved with addition of late MRI examination both in the determination of NDI or death (receiver operating characteristic area under the curve .809 vs .826) and significant gross motor impairment or death (.885 vs .908).