Discussion
TH is fundamental to the management of NE in the newborn. However, in adults, TH is not recommended in the context of sepsis as it is associated with increased morbidity18 and mortality.19 Importantly, a significant proportion of neonates treated with TH also have EOS.16 17 In the current analysis, approximately 2.5% of the infants receiving TH were culture positive. Importantly, mortality rates were not higher, but newborns with NE and sepsis did require more intensive care support with a significantly greater need for mechanical ventilation, HFOV and iNO for persistent pulmonary hypertension (PPHN). This suggested a greater illness severity during the period of TH and was associated with a significantly longer length of hospital stay in survivors. A greater proportion of the infants with sepsis were discharged home with gastrostomy tube feeding (5% vs 13%), which may suggest more neurological impairment, though results did not quite meet significance.
The literature on the combined effect of NE and sepsis in newborns who received TH is conflicting. The current data are consistent with a European study of infants with perinatal asphyxia and probable or confirmed EOS, which did not identify an increased risk of mortality.20 In contrast, a systematic review of infants with invasive group B Streptococcus infection treated with TH found an increased risk of mortality (risk ratio 2.07, 95% CI 1.47 to 2.91).21 However, this review included infants from several countries including low-income settings where patients may have limited access to similar levels of neonatal intensive care.
The overall mortality rate in the current study was 13% and did not differ between the NE group and the NE and sepsis group. This is higher than reported mortality rates for term newborns with EOS alone, which are typically reported to be approximately 1.5%–3%22–24 highlighting the significant burden of concurrent NE secondary to perinatal asphyxia. Overall, the rate of mortality for our cohort was comparable with the 10.9% reported in a review of infants who received TH in England and Wales.25 Over the last decade, the mortality for infants treated with TH has improved from approximately 27% while the proportion of infants cooled with culture-proven sepsis has remained similar.16 This may reflect changes in clinical practice with more infants with less severe NE receiving treatment with TH,17 advances in intensive care and earlier recognition and treatment of NE.
In the current study, infants with NE and sepsis were more likely to require mechanical ventilation compared with those without sepsis. Further, the duration of mechanical ventilation was longer and they were more likely to receive support with HFOV and iNO. It is known that infants with sepsis are at risk of PPHN, a known complication of TH.16 A recent case report of NE with sepsis treated with TH described severe PPHN with respiratory failure requiring ECMO.26 Further, the clinical course was complicated by pulmonary haemorrhage thought to be secondary to left ventricular dysfunction with asphyxia, increased pulmonary blood flow with patent ductus arteriosus, capillary dysfunction and coagulopathy.26 In our cohort, few infants were treated with ECMO with no significant difference between the groups, and pulmonary haemorrhage was not reported in either group. We do not have data on the frequency of hypotension requiring treatment as this was not collected in the ANZNN data set. However, data from animal models27 and from case reports26 suggest that the use of TH in infants with sepsis may be associated with profound hypotension refractory to medical treatment.
In considering the mechanism of neurological injury and resulting outcome, it is recognised that infection prior to acute hypoxia lowers the injury threshold exacerbating brain injury28–30 and that the extent of injury may be pathogen dependent.31 Further, animal data suggest a pre-existing infection may attenuate the therapeutic effects of cooling.27 29 30 For instance, in piglets, TH is not protective in those with Escherichia coli sepsis, based on recovery of amplitude-integrated electroencephalogram, magnetic resonance spectroscopy lactate/ N-acetylaspartate (NAA) peak area ratios in the thalamus and white matter, and histological cell death.27 TH has also been reported to be ineffective at suppressing microglial activation, proinflammatory cytokine production and astrogliosis in preclinical models of infection and acute hypoxia.27 32 In human infants, a report of term newborns with encephalopathy treated with TH reviewed for evidence of maternal chorioamnionitis and infant infection reported that newborns with NE and early bacteraemia had significant rates of moderate to severe basal ganglia or watershed injury on MRI, and a lower cognitive score on Bayley-III developmental assessment at 30 months of age.33 A further study of placental pathology has also reported chorioamnionitis with fetal vasculitis and chorionic plate meconium to be associated with brain injury on MRI.34 Finally, a recent paper reports an association between adverse short-term MRI outcome and both higher interleukin 6 (IL-6) before TH initiation and higher C-reactive protein (CRP) levels during and after TH.35 These observations have driven research into adjuvant therapies with immunomodulatory potential for infants with acute hypoxic injury including erythropoietin, melatonin, magnesium, xenon, mesenchymal stem cells, steroids and anti-inflammatory cytokines, among others.36 Although further validation is needed, work in this area is important.
Earlier, more accurate diagnosis is also a research goal as clinical signs of asphyxia may be indistinguishable from sepsis in a newborn infant.37 A positive blood culture is the gold standard for diagnosis, but results are rarely available prior to 6 hours of age.38 Currently used biomarkers include a complete blood count, though diagnostic performance is poor prior to 4 hours of age and acute hypoxia alters neutrophil dynamics.39 Acute phase reactants such as CRP are non-specific with both infection and inflammation secondary to hypoxic injury associated with increased levels, and additionally the rise may be delayed.39 Procalcitonin is more sensitive for bacterial infection, though it requires further study for its use in neonatal EOS.40 Multiplex bacterial PCR enables more rapid detection of bacterial pathogens, though it is not yet widely available.41 Other biomarkers currently under investigation include IL-6, IL-8, serum amyloid A, neutrophil CD64 and cytokines such as tumour necrosis factor.42
A major strength of the current study is the reliability of the ANZNN data set with data checked for discrepancy by the local centre prior to submission and again on acceptance by ANZNN. The 5-year period studied and the availability of data from across ANZNN also ensured an adequate sample size based on a reasonable number of infants with culture-proven sepsis. The a priori decision to only include newborns in the NE and sepsis group who had blood culture-positive sepsis is both a strength and a limitation. While controlling for the uncertainty of ‘suspected sepsis’, we acknowledge that blood cultures may be falsely negative due to an inadequate sample volume or prior exposure to antibiotics. The current cohort included all infants treated with TH but the severity of NE was not defined, which may limit the interpretation of the results. An additional limitation is that ANZNN collects a predetermined limited data set without the ability to go back to source documents, which limits the reporting on hypotension and treatment, MRI and standardised severity of illness score for all infants. As a result, the impact of severity of NE in those newborns also diagnosed with blood culture-positive sepsis, if any, cannot be assessed. A further consideration is the impact of missing data from the ANZNN data set. This is particularly relevant for the logistic regression analysis of factors predicting death. While all the data were available for ventilation, sex and sepsis, 1% of data regarding 5 min Apgar were missing for the cooled no sepsis group and 38% and 60% of cord lactate data were missing for the cooled no sepsis group and cooled sepsis group, respectively. However, no significant differences between those infants with and without cord lactate data in either group were seen (data not shown), suggesting that the data were missing completely at random. As a result, a complete case analysis approach rather than multiple imputation was chosen. Finally, the focus of the current study was mortality and short-term outcomes in the immediate newborn period. The impact of TH on longer term neurodevelopmental outcomes in those newborns with NE and sepsis was beyond the scope of this study but is a significant knowledge gap that requires further investigation.