ObjectivesTo examine if therapeutic hypothermia reduces the composite outcome of death, moderate or severe disability at 18 months or more after mild neonatal encephalopathy (NE).Data sourceMEDLINE, Cochrane database, Scopus and ISI Web of Knowledge databases, using ‘hypoxic ischaemic encephalopathy’, ‘newborn’ and ‘hypothermia’, and ‘clinical trials’ as medical subject headings and terms. Manual search of the reference lists of all eligible articles and major review articles and additional data from the corresponding authors of selected articles.Study selectionRandomised and quasirandomised controlled trials comparing therapeutic hypothermia with usual care.Data extractionSafety and efficacy data extracted independently by two reviewers and analysed.ResultsWe included the data on 117 babies with mild NE inadvertently recruited to five cooling trials (two whole-body cooling and three selective head cooling) of moderate and severe NE, in the meta-analysis. Adverse outcomes occurred in 11/56 (19.6%) of the cooled babies and 12/61 (19.7%) of the usual care babies (risk ratio 1.11 (95% CIs 0.55 to 2.25)).ConclusionsCurrent evidence is insufficient to recommend routine therapeutic hypothermia for babies with mild encephalopathy and significant benefits or harm cannot be excluded.