Abstract: Summary Background Cerebral hypothermia can improve outcome of experimental perinatal hypoxia-ischaemia. We did a multicentre randomised controlled trial to find out if delayed head cooling can improve neurodevelopmental outcome in babies with neonatal encephalopathy. Methods 234 term infants with moderate to severe neonatal encephalopathy and abnormal amplitude integrated electroencephalography (aEEG) were randomly assigned to either head cooling for 72 h, within 6 h of birth, with rectal temperature maintained at 34–35°C (n=116), or conventional care (n=118). Primary outcome was death or severe disability at 18 months. Analysis was by intention to treat. We examined in two predefined subgroup analyses the effect of hypothermia in babies with the most severe aEEG changes before randomisation—ie, severe loss of background amplitude, and seizures—and those with less severe changes. Findings In 16 babies, follow-up data were not available. Thus in 218 infants (93%), 73/110 (66%) allocated conventional care and 59/108 (55%) assigned head cooling died or had severe disability at 18 months (odds ratio 0·61; 95% CI 0·34–1·09, p=0·1). After adjustment for the severity of aEEG changes with a logistic regression model, the odds ratio for hypothermia treatment was 0·57 (0·32–1·01, p=0·05). No difference was noted in the frequency of clinically important complications. Predefined subgroup analysis suggested that head cooling had no effect in infants with the most severe aEEG changes (n=46, 1·8; 0·49–6·4, p=0·51), but was beneficial in infants with less severe aEEG changes (n= 172, 0·42; 0·22–0·80, p=0·009). Interpretation These data suggest that although induced head cooling is not protective in a mixed population of infants with neonatal encephalopathy, it could safely improve survival without severe neurodevelopmental disability in infants with less severe aEEG changes.