MRI and withdrawal of life support from newborn infants with hypoxic-ischemic encephalopathy

The majority of deaths in infants with hypoxic-ischemic encephalopathy (HIE) follow decisions to withdraw life-sustaining treatment. Clinicians use prognostic tests including MRI to help determine prognosis and decide whether to consider treatment withdrawal. A recently published meta-analysis provided valuable information on the prognostic utility of magnetic resonance (MR) biomarkers in HIE and suggested, in particular, that proton MR spectroscopy is the most accurate predictor of neurodevelopmental outcome. How should this evidence influence treatment-limitation decisions? In this article I outline serious limitations in existing prognostic studies of HIE, including small sample size, selection bias, vague and overly inclusive outcome assessment, and potential self-fulfilling prophecies. Such limitations make it difficult to answer the most important prognostic question. Reanalysis of published data reveals that severe abnormalities on conventional MRI in the first week have a sensitivity of 71% (95% confidence interval: 59%-91%) and specificity of 84% (95% confidence interval: 68%-93%) for very adverse outcome in infants with moderate encephalopathy. On current evidence, MR biomarkers alone are not sufficiently accurate to direct treatment-limitation decisions. Although there may be a role for using MRI or MR spectroscopy in combination with other prognostic markers to identify infants with very adverse outcome, it is not possible from meta-analysis to define this group clearly. There is an urgent need for improved prognostic research into HIE.