Benefits and risks of therapeutic hypothermia for hypoxic-ischemic encephalopathy in late preterm infants

Abstract

Abstract: Therapeutic hypothermia (TH) is standard care for term neonates with moderate to severe hypoxic-ischemic encephalopathy (HIE), but its use in preterm infants 33–35 weeks’ gestational age (GA) remains controversial. This review article summarizes the biological rationale, clinical evidence, and real-world experience supporting or challenging TH in this population. Preclinical models show neuroprotective effects of TH at developmental stages equivalent to late preterm infants. Retrospective studies suggest feasibility but report higher complication rates, particularly at 34 weeks. We critically evaluate the only randomized controlled trial (RCT) to date, which reported no benefit and possible harm with TH in 33–35 weeks’ GA infants. However, this study had important limitations, including baseline imbalances, limited stratification by GA and encephalopathy severity, and lack of neuroimaging or EEG data. A recent international survey of 88 centers reveals heterogeneous practices, with many continuing TH at 34–35 weeks’ GA despite the trial’s findings. Real-world experience from 22 centers shows lower mortality than reported in the RCT. TH may remain appropriate for select 35 weeks’ GA infants, but routine use in 34 weeks’ GA and earlier infants should be limited to research settings. Future studies should stratify by GA and include standardized neurological assessments to inform practice. Impact of This Review: Summarizes preclinical and retrospective clinical data supporting the biological plausibility and feasibility of therapeutic hypothermia (TH) in late preterm infants. Critically examines the limitations of the only RCT of TH in this population and explains why its findings should not be generalized to all 33–35 weeks’ GA infants. Highlights real-world evidence showing lower mortality and supports continued TH use at 35 weeks’ GA while urging caution at 34 weeks or below pending further data.