Objective To determine precooling attributes possibly predicting death in infants with hypoxic-ischaemic encephalopathy (HIE) despite therapeutic cooling.
Methods Eighty-five consecutive infants of ≥36 weeks' gestation who received cooling for HIE were reviewed. Logistic regression analysis was performed using precooling clinical and laboratory variables with death related to HIE during the first 9 months of life as the primary outcome.
Results Thirteen (15%) of the 85 infants died during 9–18 months of follow-up despite cooling. 27 of the 85 were asystolic at birth but only 12 had Apgar scores of zero at both 5 and 10 min. Univariate analysis identified Apgar scores of zero at 5 and 10 min, pH <6.7, base deficit >22 mmol/l, and absent spontaneous movement as significantly associated with death during the first 9 months despite cooling. On multivariate analysis, only the Apgar score of zero at 10 min (p<0.001, OR 51.7, 95% CI 9.9 to 269.5) remained significantly associated with the primary outcome of death from HIE. Of the 12 infants who were asystolic at and beyond 10 min of life, nine died from HIE, two had spastic quadriparesis and global delay at 18–24 months, and one had extensive encephalomalacia on brain MRI during follow-up.
Conclusions Of the selected precooling variables, only the 10 min Apgar score is independently associated with death despite therapeutic cooling in infants with HIE. Infants who remain asystolic at 10 min and beyond are unlikely to survive despite cooling, and the rare survivor is likely to have severe disability.
Statistics from Altmetric.com
During discussion of outcomes with parents before an infant with asphyxia is started on a cooling protocol, the parents often ask if their infant is going to survive. Therapeutic cooling in asphyxiated newborns improves survival, but some cases with severe asphyxia “may never be treatable”.1 2 Indeed, in a meta-analysis of the Cool Cap, National Institute of Child Health and Human Development (NICHD) and TOBY trials, Azzopardi and colleagues3 recently reported that therapeutic hypothermia was less effective in infants with severe encephalopathy. At randomisation these infants were a heterogeneous group with variable response to therapeutic hypothermia, indicating that some were not ‘treatable’ because injury was too severe. Brain injury too severe to respond to treatment specifically refers to brain injury in apparently stillborn full term infants who remain asystolic beyond 10 min despite resuscitation efforts. As these newborns rarely survive, and the rare survivor has severe disability, the current Neonatal Resuscitation Program (NRP) of the American Academy of Pediatrics states that discontinuation of resuscitation efforts may be appropriate for these infants if there are no other reversible causes of newborn compromise.4,–,7 However, it is not known if the outcome for apparently stillborn infants who remain asystolic at 10 min despite resuscitation efforts would be any different following therapeutic hypothermia.
What is already known on this topic
▶ Therapeutic hypothermia in asphyxiated newborns reduces mortality without increasing major disability in survivors.
▶ Given the underlying pathophysiology of hypoxic-ischaemic encephalopathy, it may be impossible to rescue the brains of all affected infants.
What this study adds
▶ Apgar score of zero at 10 min and beyond helps in predicting death despite therapeutic cooling in infants with hypoxic-ischaemic encephalopathy.
▶ Neuroprotection with cooling may not prevent neonatal death or neurodevelopmental impairment in apparently stillborn infants who remain asystolic at 10 min and beyond.
Because cooling needs to be initiated within 6 h after birth to be effective, it would be useful to be able to evaluate the utility of hypothermia in infants with hypoxic-ischaemic encephalopathy (HIE), specifically in apparently stillborn infants who remain asystolic at 10 min, for better counselling regarding outcomes and to avoid offering false hope to parents.
Hence, the objective of this study was to determine the precooling attributes predicting death during the first 9 months of life in asphyxiated infants with HIE despite therapeutic hypothermia. A secondary goal was to determine if outcomes for apparently stillborn infants who remain asystolic at 10 min and beyond were different with and without therapeutic hypothermia than what has been historically true for such infants.5,–,7
Patients and methods
We reviewed the medical records of all asphyxiated newborns who received cooling at the University of Michigan for treatment of suspected hypoxic-ischaemic brain injury between December 2003 and February 2009. The University of Michigan was a participating centre in the Cool Cap (selective head cooling (SHC)) trial.8 Once the Cool Cap trial ended, SHC was offered for hypothermic neuroprotection under an FDA-approved ‘continued access protocol’ to infants ≥36 weeks' gestation with moderate or severe HIE using the same stepwise recruitment criteria as in the original Cool Cap trial. However, subsequently we have gradually switched to whole body cooling (WBC).9 Early in our use of WBC, infants with moderate or severe HIE admitted to our neonatal intensive care unit underwent WBC if they fulfilled the same clinical and laboratory eligibility criteria of the Cool Cap protocol but had a ‘non-qualifying’ amplitude-integrated electroencephalogram (aEEG),10 or if the Cool Cap devices were unavailable. More recently, infants at risk of significant hypoxic-ischaemic brain injury are only offered WBC using the recruitment criteria of the NICHD Neonatal Research Network WBC protocol.9 Clinical monitoring and treatment of the infants during cooling, whether SHC or WBC, were performed according to the Cool Cap and NICHD protocols.8 9 Written informed parental consent was obtained before cooling for all infants. This retrospective review of data was approved by the Institutional Review Board at the University of Michigan.
Stepwise logistic regression analysis was performed to examine which of the selected precooling (before 6 h of age) clinical and laboratory variables predict death related to asphyxial brain injury despite therapeutic hypothermia during the neonatal period and 9 months of follow-up. Unlike other cooling trials,8 9 11 we focused on death despite cooling as the primary outcome instead of the combined rate of death and severe disability because infant survival is the main concern of parents when an infant is started on a cooling protocol and because a recent meta-analysis of cooling trials showed that cooling reduces mortality without increasing major disability in survivors.12
The selected precooling covariates for this exploratory analysis included: 5 and 10 min Apgar scores of zero; pH of <6.7; base deficit of >18.5 mmol/l, or base deficit of >22 mmol/l in cord blood or during the first 1 h after birth; presence or absence of clinical seizure; presence or absence of specific components of early neurological examination, including flaccidity, decerebrate posturing, spontaneous activity and suck/gag reflexes; and presence or absence of an intrapartum sentinel event (abruption, cord accident and/or ruptured uterus, and meconium stained amniotic fluid with fetal bradycardia). This selection was based on previous observations that not all components of early neurological examination indicate prognosis; the selected covariates were similar to variables previously noted to predict severe disability or death in infants with HIE despite therapeutic hypothermia.13 Compared to encephalopathy grade which was used in the secondary analysis of the Cool Cap trial as a determinant of outcome,14 the specific components of early neurological examination we used were reportedly more predictive of death or disability in infants with HIE despite therapeutic hypothermia.13 Rare prenatal/intrapartum events were not entered in the univariate analysis of the logistic regression model.
Data collection was concurrent with the cooling protocol and included variables related to prenatal/intrapartum events (birth weight, gestational age, mode of delivery, meconium staining of amniotic fluid, outborn/inborn, home birth, chronic hypertension/pre-eclampsia/eclampsia, placental problems (abruption/previa), progress of labour and shoulder dystocia, uterine rupture, cord accident (cord prolapse, cord avulsion, vasa previa), maternal fever, fetal bradycardia and gender of the infant), resuscitation (asystole at birth; Apgar scores at 1, 5 and 10 min, and beyond if available; time of first audible heart beat; first blood gas pH and base deficit within 60 min of birth; details of chest compression, intubation and bag/mask ventilation; and medications for resuscitation) and postnatal events (presence of moderate or severe clinical encephalopathy according to modified Sarnat and Sarnat criteria before 6 h of age, including tone, posture, spontaneous activity and suck/gag reflexes; aEEG recordings, if done, prior to cooling; and type of hypothermic neuroprotection (SHC or WBC) and clinical seizures before cooling).
Given that point estimates for the neuroprotective effects of cooling were essentially identical between the large SHC and WBC trials,12 the data for our study population receiving either SHC or WBC were merged for analysis.
Eighty-six consecutive cases of asphyxiated infants of ≥36 weeks' gestation who received therapeutic hypothermia for moderate (stage 2) or severe (stage 3) HIE between December 2003 and February 2009 were reviewed. Forty-five infants received SHC, 39 infants received WBC, and two additional infants initially started on SHC were subsequently switched to WBC because the SHC device malfunctioned. One infant who received WBC but was subsequently diagnosed with congenital myotonic dystrophy, was excluded from the final analysis.
The neonatal and obstetric characteristics of the final study population of 85 cooled infants are shown in table 1. During follow-up with outcome to at least 9 months (as the last infant in this series received cooling only 9 months before the time of writing) and a maximum of 18–24 months of age, 13 (15.3%) of the 85 asphyxiated infants died from HIE despite therapeutic hypothermia. Ten infants (12%) with HIE died from multiorgan dysfunction during the neonatal period (the first 4 weeks of life) and three more died during late infancy from complications related to HIE (two infants at 8 months and the other at 12 months of age).
Univariate analysis, with outcome to at least 9 months of follow-up, identified an Apgar score of zero at 5 min, an Apgar score of zero at 10 min, pH of <6.7, base deficit of >22 mmol/l, and absent spontaneous movement as being significantly associated with death during the neonatal period and the first 9 months of life (table 2). On multivariate analysis, only an Apgar score of zero at 10 min (p<0.001, OR 32.7, 95% CI 6.4 to 166.5) independently predicted death during the first 4 weeks of life. When the logistic regression analysis was repeated with outcome to at least 9 months, results were similar and again, only an Apgar score of zero at 10 min remained significantly associated with death (p<0.001, OR 51.7, 95% CI 9.9 to 269.5), with a sensitivity of 75%, specificity of 96%, positive predictive value (PPV) of 75%, negative predictive value (NPV) of 96%, positive likelihood ratio of 18.8, and a negative likelihood ratio of 0.26 (table 3).
Twenty-seven of the 85 infants were asystolic at birth with Apgar scores of zero at 1 min; 12 of these infants had Apgar scores of 0 at both 5 and 10 min. Table 4 shows the clinical details of infants with an Apgar score of zero at 10 min and beyond who were resuscitated and transferred to our unit for therapeutic hypothermia. Of the 12 infants who were asystolic beyond 10 min of life, nine infants (75%) died from HIE and related complications (seven during the neonatal period, and two at 8–12 months of age), two infants had spastic quadriparesis with severe global delay at 18–24 months' follow-up, and one infant had extensive encephalomalacia on brain MRI at 2 months of age, with microcephaly and seizures but no appreciable tone abnormality at 6 months' follow-up. During follow-up with outcome to at least 9 months and a maximum of 18–24 months of age, only four (5%) of the 73 cooled infants who were not asystolic at 10 min died: three died from multiorgan dysfunction related to HIE (one on day 2 while being cooled, one on day 6 after receiving 72 h of cooling, and one at 8 months of age), and one died on day 27 from complications of extracorporeal membrane oxygenation (ECMO) for severe persistent pulmonary hypertension of the newborn (PPHN) complicated by renal failure. None of the four infants were asystolic at birth.
At the time of writing, clinical outcome data at 18–24 months of follow-up were available for 62 (86%) of the 72 surviving infants. In the other 10 surviving infants, clinical outcome data were available at 12–18 months of age for nine patients, and at 9–12 months of age for one patient. One of these 10 infants was asystolic at 10 min of age (patient number 11 in table 4), while five of the other nine infants who were not asystolic at birth or beyond, had normal postcooling brain MRIs. Neurodevelopmental follow-up is continuing in the other four infants.
Therapeutic hypothermia is beneficial for term newborns with HIE.12 However, data from randomised control trials suggest that cooling may not be protective in infants with severe brain injury.15 Two studies examined the data from the Cool Cap and the NICHD WBC trials to determine which infants were and were not likely to benefit from cooling.13 14 The authors of the Cool Cap trial noted that outcomes after hypothermic treatment were strongly influenced by the severity of HIE including aEEG abnormality.14 Similarly, investigators from the NICHD WBC trial used a scoring system or a decision tree scheme incorporating precooling clinical and laboratory variables to determine which infants were unlikely to benefit from cooling.13 Interestingly, none of these trials incorporated an Apgar score of zero at 10 min and beyond as one of the precooling variables in the analysis. Our observational study in a large cohort of infants, although non-randomised and from a single centre, demonstrates that of all the variables, only the 10 min Apgar score is useful for predicting death in infants with HIE despite therapeutic hypothermia. The results of the secondary outcome data indicate that infants who are apparently stillborn and remain asystolic at 10 min and beyond are very unlikely to survive despite therapeutic hypothermia (with the odds of death increasing over time, as is evident from the increase in the odds ratio of 32.7 during the neonatal period to 51.7 at 9 months), and the rare survivor is likely to have severe disability. This finding is consistent with follow-up studies of HIE raising concerns that outcome worsens with longer follow-up. This finding is also consistent with limited but consistent prehypothermia data that this group of infants either dies or survives with severe disability.5,–,7 We believe that this group of infants represents ‘untreatable’ cases for whom cooling may not be of benefit. Moreover, it is important to note that in this study the NPV of an Apgar score of zero at 10 min for predicting neonatal death despite therapeutic hypothermia was 96%. A high NPV indicates that if the heart rate returns even in an apparently stillborn infant and the Apgar score is more than zero at 10 min, the infant is unlikely to die if treated with cooling. Furthermore, specificity of 96% indicates that the return of a heart beat by 10 min correctly identifies 96% of likely survivors if treated with therapeutic hypothermia. This information may be useful in discussion with parents concerning the risk of death in their infant and in risk stratification of infants with HIE in future cooling trials.13
Several factors may contribute to the poor outcome despite therapeutic hypothermia in apparently stillborn infants who remain asystolic at 10 min and beyond. One factor is probably related to failure of initial recovery of the cerebral oxidative metabolism because brain injury was too severe or because these infants have already entered the secondary phase of energy failure.15 It is also possible that more severe insults may had resulted in rapid progression of delayed neuronal death with corresponding reduced effectiveness of hypothermia therapy.15
It is evident from our study that contrary to NRP guidelines,4 resuscitation efforts are being continued in most referring centres for a significant length of time even in infants who are asystolic at birth and remain asystolic at 10 min and beyond despite resuscitative efforts. Our data on secondary outcomes indicate that after 10 min of asystole, therapeutic hypothermia may not prevent death or neurodevelopmental impairment. However, the number of infants who were asystolic at 10 min and beyond was small, and favourable outcomes with cooling might be seen in a larger cohort. Also, an Apgar score of zero assigned at 10 min is not the same as asystole after 10 min of adequate resuscitation. Resuscitation efforts are frequently inadequate,16 and assignment of an Apgar score also may be inaccurate as subjective components can influence assessment of heart rate. For example, cord pulsations are not always felt17 and heart beats may be faint or almost inaudible if an infant is in shock18 or hydropic with or without pericardial effusion. Hence, the findings of this study should not be used for decisions to withhold or discontinue resuscitation efforts.
Our single-centre retrospective analysis has other limitations. Results of our logistic regression analysis indicating asystole at 10 min as the single most predictive factor for death may be less accurate in a different population or cohort. Furthermore, outcomes to 18 months of age and beyond were not available in 10 surviving infants. Long term follow-up is important, particularly in light of a recent report describing mild or no disability on longer follow-up in nearly 25% of infants with an Apgar score of zero at 10 min enrolled in the NICHD cooling trial.19 Five of the 10 study infants with shorter follow-up (to at least 9 months of age) had normal postcooling brain MRI and are unlikely to die from HIE before they reach 18 months of age. However, long-term monitoring is continuing for these infants. Also, similar to other clinical trials predicting outcome,13 14 only selected precooling variables were analysed, so other potentially important variables may have been omitted. However, no previous model for determining the prognosis of asphyxiated infants undergoing therapeutic cooling had a 100% correct prediction rate.13 14
Therefore, despite all the limitations, this observational study is an important reminder to clinicians caring for infants with HIE of the value of recording the 10 min Apgar score correctly, as it helps in predicting death in infants with HIE despite cooling. Furthermore, because of calls to offer cooling to every patient with asphyxial brain injury,20 clinicians should be mindful that neuroprotection with cooling may not prevent neonatal death or neurodevelopmental impairment in apparently stillborn infants who remain asystolic at 10 min and beyond.
Competing interests None.
Ethics approval This study was conducted with the approval of the University of Michigan Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.