Discussion

Here we report the brain injury and neurodevelopmental outcomes after premature cessation of cooling therapy in a small series of babies with mild encephalopathy.

Discontinuation of cooling therapy due to adverse events or for palliative care is a well-accepted clinical practice, hence we specifically excluded such cases. Discontinuation of cooling in our cohort was driven by a view that therapeutic hypothermia was no longer needed as the baby had already recovered clinically. Uncertainty also existed about whether the infant had fulfilled the cooling criteria in first place (moderate or severe encephalopathy). This often occurs when passive cooling therapy is initiated at a local hospital prior to the referral, followed by a detailed neurological evaluation at the cooling centre, during which the baby no longer exhibits signs of encephalopathy.

It is unclear if brain injury might have been averted if cooling therapy was given up to 72 hours. Walsh and colleagues from Boston Children Hospital have recently reported brain injury on MRI in 54% of the babies with mild neonatal encephalopathy, despite complete cooling therapy.5 Nevertheless, in preclinical models, brief hypothermia after hypoxia-ischaemia only postpones brain injury and does not provide any long-lasting neuroprotection.6 Furthermore, short cooling therapy (48 hours) offers less neuroprotection than standard cooling (72 hours) in a fetal sheep model of perinatal encephalopathy.7

Inflammatory markers are commonly elevated in neonatal encephalopathy even without coexistent infection,8 and only one baby in our series had proven blood stream positive sepsis (group B streptococci). Preclinical data suggest that hypothermic neuroprotection may be lost with coexistent Gram-negative infections, but not with Gram-positive infections.9 Hence, any lack of neuroprotection in our cohort is unlikely to be due to coexistent perinatal infection.

There are three major limitations of our study. First, the reported incidence of premature cessation of cooling in our study (24%) is likely to be an underestimate, when compared with actual clinical practice. Our babies were part of a multicentre study, and hence may have had closer monitoring. A recent survey in Australia reported a 20% (21/104) incidence of premature cessation of cooling therapy (due to clinical improvement (n=16) or lack of qualification (n=5) in retrospect) in 104 babies who were cooled despite not meeting the cooling criteria.1 Cooling therapy was also prematurely discontinued in 2% (2/103) of the babies who had originally met the cooling criteria (moderate or severe encephalopathy), due to an apparent clinical improvement. Neuroimaging or neurodevelopmental outcome data were not available in this cohort, nevertheless the authors do raise concerns about such practices. Second, although none of our cases seemed to have met the original cooling criteria and had a rapid clinical recovery, the high incidence of brain injury and the adverse neurodevelopmental outcomes in these babies is worrying. This questions whether these babies may have in fact suffered moderate (not mild) injury, and reflects the practical challenges in accurately identifying the severity of encephalopathy within 6 hours of birth. Finally, we have not compared the outcomes of complete versus incomplete cooling therapy in babies with mild encephalopathy.

The long-held view of uniformly good outcomes in mild encephalopathy is being increasingly challenged in more recent studies. Growing evidence suggests that cognitive impairments and behavioural problems during childhood following mild neonatal encephalopathy are closer than previously thought to those of children with moderate encephalopathy.10 Although in three cases the MRI abnormalities did not correlate with 2-year outcomes in our series, they could be associated with adverse outcomes during childhood.

Although classical secondary energy failure may not occur in mild encephalopathy, it is possible that perinatal hypoxic injury is a continuum. For example, perinatal asphyxia without encephalopathy leads to a reduction in IQ compared with healthy peers, and mild, moderate and severe encephalopathy results in increasing severity of long-term adverse outcomes. Hypothermia is reported to have a marked neuroprotective effect in a recently described preclinical model of mild encephalopathy, where the predominant injury was in the white matter.11 Preclinical studies also report that brain cooling in the absence of perinatal asphyxia may increase apoptosis.12

The data presented in this case series suggest that despite clinician perception of clinical recovery, the premature cessation of cooling therapy in babies with mild neonatal encephalopathy does not exclude residual brain injury and adverse long-term neurodevelopmental outcomes. The creeping introduction of non-standard cooling practices, constantly challenged by residual safety concerns, can only be resolved by large pragmatic trials of cooling in mild encephalopathy. In addition to neurodevelopmental outcome assessments, these studies should also examine the cost–benefit implications, as cooling therapy is expensive and may prolong hospital stay. Such trials are currently in development in the UK and USA.