|Home | About | Journals | Submit | Contact Us | Français|
Moderate to severe hypoxic–ischemic injury in newborn infants, manifested as encephalopathy immediately or within hours after birth, is associated with a high risk of either death or a lifetime with disability. In recent multicenter clinical trials, hypothermia initiated within the first 6 postnatal hours has emerged as a therapy that reduces the risk of death or impairment among infants with hypoxic–ischemic encephalopathy. Prior to hypothermia, no therapies directly targeting neonatal encephalopathy secondary to hypoxic–ischemic injury had convincing evidence of efficacy. Hypothermia therapy is now becoming increasingly available at tertiary centers. Despite the deserved enthusiasm for hypothermia, obstetric and neonatology caregivers, as well as society at large, must be reminded that in the clinical trials more than 40% of cooled infants died or survived with impairment. Although hypothermia is an evidence-based therapy, additional discoveries are needed to further improve outcome after HIE. In this article, we briefly present the epidemiology of neonatal encephalopathy due to hypoxic–ischemic injury, describe the rationale for the use of hypothermia therapy for hypoxic–ischemic encephalopathy, and present results of the clinical trials that have demonstrated the efficacy of hypothermia. We also present findings noted during and after these trials that will guide care and direct research for this devastating problem.
Neonatal encephalopathy, or a persistently abnormal neurologic examination that includes one or more of the following: altered consciousness, abnormal muscle tone or reflexes, altered respirations, and sometimes seizures in the first postnatal hours to days – can be caused by a number of etiologies, including acute perinatal asphyxia, intracranial bleeding and stroke, injuries secondary to birth trauma, infections, metabolic disorders including hypoglycemia, and congenital brain abnormalities. Encephalopathy due to hypoxic–ischemic injury – hypoxic–ischemic encephalopathy (HIE) – is defined as brain injury caused by the combination of inadequate blood flow and oxygen delivery to the brain. Attributing neonatal encephalopathy to perinatal hypoxic–ischemic injury requires combinations of parameters indicative of metabolic acidosis in the first postnatal hours with low cord pH (<7.0), base deficit of over 12, and evidence of a need for respiratory support also starting in the first minutes, with low Apgar scores at and beyond 5 min. The American College of Obstetrics and Gynecology maintains that only specific types of long-term neurologic injury, including specific types of cerebral palsy, can be connected to perinatal hypoxic–ischemic injury .
In a recent review, Graham et al. found that HIE occurs in an estimated 2.5 (1.2–7.7) of every 1000 term births in high-income countries . In less affluent countries, the incidence is reported to be higher; in one example, the incidence was 26 per 1000 . Worldwide, approximately a quarter of all neonatal (first postnatal month) deaths are attributed to perinatal asphyxia, a leading cause of HIE [4,5]. Cerebral palsy is a significant morbidity associated with HIE, but the percentage of cerebral palsy cases associated with prior HIE is estimated to be 15% . Children born at term who develop cerebral palsy following newborn encephalopathy have a poorer prognosis (twice as likely to have a severe composite disability score and three-times more likely to die in the first 5 postnatal years) than those with cerebral palsy who were not encephalopathic in the first week of life .
Clinicians caring for neonates with HIE generally describe their clinical appearance in the first postnatal days in terms of Sarnat scores, or slightly modified variants of these scores. The scores are based on serial neurobehavioral evaluations (TABLE 1) . Neonates with mild encephalopathy (Sarnat I) do not have an increased risk of motor or cognitive deficits. Neonates with persistent severe encephalopathy (Sarnat III) have a risk of death, and an increased risk of cerebral palsy and mental retardation among survivors that approaches 100%. Neonates with moderate encephalopathy (Sarnat II) are less likely to have severe impairments, but may still have significant deficits, memory impairment, visual motor or visual perceptive dysfunction, increased hyperactivity, and delayed school readiness [8,9]. Adding serial MRI neuroimaging, inclusive of diffusion-weighted imaging, which measures the diffusion of water in tissues (less diffusion is proportional to more injury) to demonstrate evolving pathology in the first postnatal months can add to the predictive value of the early neonatal period exams . Neuromonitoring with conventional EEG and amplitude-integrated EEG (aEEG; discussed later) can also provide additional information regarding current status and can be helpful in predicting long-term outcome [11–13]. Combinations of all three, serial physical exam assessment, MRI, and EEG and aEEG, are likely to be most useful for prognosis, but the availability of EEG and MRI with diffusion-weighted imaging capability for neonates may be limited in many clinical settings.
The pathophysiology of brain injury secondary to hypoxic–ischemic injury can be simplified into two phases of pathologic events that culminate, after weeks, in brain injury, and after months to years, in measureable neurodevelopmental dysfunction. These phases are primary and secondary energy failure, with a latency period of varying length inbetween (FIGURE 1). This pathophysiolgic model for hypoxic–ischemic brain injury is based on characteristics of the temporal sequence of cerebral energy state in timed injury newborn animal models . Primary energy failure is characterized by reductions in cerebral blood flow and, consequently, delivery of oxygen and substrates to brain tissue [14,15]. High-energy phosphorylated compounds are reduced, and brain tissue acidosis is prominent. This phase is a prerequisite for all subsequent deleterious events. Primary energy failure is associated with acute intracellular derangements, including loss of membrane ionic homeostasis, release and blocked reuptake of excitatory neurotransmitters, defective osmoregulation and inhibition of protein synthesis . Excessive stimulation of excitatory neurotransmitter receptors and loss of ionic homeostasis mediate an increase in intracellular calcium and osmotic dysregulation. Elevation in intracellular calcium triggers a number of destructive pathways by activating lipases, proteases and endonucleases, ultimately resulting in acute cell death and necrosis .
Resolution of hypoxia–ischemia within a specific time interval reverses the primary energy failure fall in high-energy phosphorylated metabolites and intracellular pH, and promotes recycling of neurotransmitters. The duration of time for hypoxia–ischemia to be successfully reversed and promote recovery is likely influenced by maturation, preconditioning events, substrate availability, body temperature and simultaneous disease processes, and possibly individual genetic variations. Following primary energy failure, cerebral metabolism may recover following reperfusion–reoxygenation, only to deteriorate in a secondary energy failure. This secondary energy failure differs from primary energy failure in that declines in phosphocreatine and ATP are not accompanied by brain acidosis . The presence and severity of secondary energy failure depends on the extent of primary energy failure. The pathogenesis of secondary energy failure is not as well understood as primary energy failure, but likely involves multiple pathophysiologic processes, including the accumulation of excitatory neurotransmitters such as glutamate, oxidative injury, anerobic metabolism and free radical production from the synthesis of xanthenes and uric acid, inflammation and, ultimately, an untimely initiation of apoptosis in many cells. In addition to classic injury concepts, in the newborn, secondary energy failure also leads to altered growth factor and protein synthesis, which ultimately adds the ongoing insult of altered brain development to the established hypoxic–ischemic injury [11,18–22]. More extensive secondary energy failure and disruption of oxidative metabolism worsens the outcome .
Much of the later injury after secondary energy failure occurs secondary to the process of apoptosis, or programmed cell death. In contrast to the cell membrane disruptions that lead to necrosis in primary energy failure, apoptosis is a more nuclear phenomenon that ultimately results in DNA fragmentation and condensation [24,25]. Apoptosis occurs as an important component of normal brain development. Cell connections and pathways are refined via apoptosis , but when the process is accelerated and apoptosis occurs in regions of the brain where it is unintended, as it does in the context of hypoxic–ischemic injury, there are significant problems. The cellular signals after hypoxic–ischemic injury accelerate apoptosis in normal brain tissue, contributing significantly to the later evolving injury noted in infants with HIE [27–29]. Although acute cellular necrosis and apoptosis both are noted post-hypoxic–ischemic injury in animal models and human infants, apoptotic cell death appears to be a more significant contributor in the developing brain of the term infant, and is the biggest problem in secondary energy failure [30,31].
The interval between primary and secondary energy failure represents a latent phase that corresponds to an optimal available therapeutic window, as we clinicians do not always know the precise time of onset and resolution of primary energy failure in newborns with HIE. Some diagnoses, such as cord prolapse or uterine rupture may suggest a time of onset of primary energy failure. In the National Institute of Child Health and Human Development (NICHD) Hypothermia trial, cord prolapse, maternal hemorrhage or uterine rupture accounted for slightly less than half of the deliveries resulting in enrolled infants . Initiation of therapies during the latent phase between resolved primary energy failure and the initial phase of secondary energy failure in perinatal animals with timed initial injury has been successful in reducing brain damage, possibly through alteration or avoidance of the secondary energy failure. These results substantiate the presence of a therapeutic window in human newborns and offered hope for postnatal intervention to improve outcome for neonatal HIE [22,26,27]. Hypothermia is one intervention that has been tested during the latency period after timed initial injury and primary energy failure .
In most cases, a neonatologist cannot address primary energy failure. Even in deliveries in which an acute event such as cord prolapse or acute abruption is noted, minutes pass before delivery occurs and neonatal resuscitation can begin, and sometimes there is uncertainty as to whether or not there was subacute compromise prior to the noted acute event, and whether or not the initial resuscitation was adequate. Owing to the uncertainty around identification of the exact timing of primary energy failure, identifying a postnatal therapy to ameliorate secondary energy failure and its subsequent significant detrimental impact on brain development is a logical target. In an anecdotal report, hypothermia had been used after cardiac arrest in four adult patients, and three had unexpected normal neurologic function afterwards . Previously, hypothermia had been used to decrease ‘brain swelling’ after injury or the removal of brain tumors . In the first report of use of hypothermia for term infants with apparent HIE, Westin reported using immersion in cold water to achieve core body temperatures of 23–25°C for no more than 20 min in six patients. Five of the infants survived to apparently normal outcomes at 6 and 30 months . This was followed by similar reports by others using similar cooling protocols [36–38]. Despite these promising results, concurrent randomized clinical trials (vs the case series’ reported for the cooling strategies for HIE) demonstrating harm from using hypothermia in preterm infants , and animal studies that failed to show benefit from hypothermia after hypoxic injury [40,41], led to questioning of the use of hypothermia for HIE, with recommendations to maintain skin-measured euthermia at 36.5°C to reduce heat loss and oxygen consumption in the immediate postnatal period for term infants with HIE [42,43]. Fortunately, investigators continued testing hypothermia in asphyxiated animal models. Interest in hypothermia as a therapeutic measure resumed in 1987, with the publication of work by Busto et al. who reported that small differences in brain temperature during ischemia had a critical effect on the extent of histologic brain injury in adult rats . This work led investigators interested in hypoxic–ischemic injury in newborns to test hypothermia in animal models of hypoxic–ischemic injury to explore mechanisms that could explain hypothermia’s anecdotal successes in newborns [14,30,45,46].
In these early ‘second era’ animal studies, investigators measured how physiologic and anatomic responses to hypoxia and ischemia were modified by moderate hypothermia (no less than 6°C below normal). Thoresen et al. used phosphorus magnetic resonance spectroscopy to demonstrate that hypoxic injury to piglets lowered the cerebral phosphocreatine:inorganic phosphate ratio to almost zero, and the nucleotide triphosphate:exchangeable phosphate pool ratio to below 30% of baseline. High-energy phosphate pools in cooled animals recovered and maintained recovery, but the injured animals maintained at normothermia experienced short-lived recovery and then a prolonged depletion . Edwards et al. demonstrated that the brains of cooled animals versus those maintained at normothermia had significant reductions in apoptosis, an established key element of secondary energy failure and ultimate cell death . As a potential explanation for the decreased apoptosis, Fukuda et al. examined the effects of hypothermia on caspase-3 activation in 7-day-old rat pups. Caspase-3 is a cysteine protease present in various kinds of cells, including neuronal cells . When caspase-3 is activated, it cleaves endogenous substrate proteins, such as inhibitory protein of caspase-activated DNase, finally causing the degradation of chromosomal DNA characteristic of apoptosis . This caspase-3 activation has been detected in hypoxic–ischemic-induced neuronal cell death . In Fukoda’s experiment, 24 h of hypothermia initiated 1 h after hypoxic–ischemic (permanent left carotid ligation) injury suppressed caspase-3 activation following injury, leading to excellent brain protection assessed at multiple time points through 168 h . Laptook et al. demonstrated that a decrease in brain temperature from normothermic values reduced brain energy utilization by approximately 5% for every 1°C cooled. This lowering of energy utilization could contribute to neuroprotection by enhancing the maintenance of high-energy ATP stores during hypoxia–ischemia [50,51]. Other neuroprotective effects of cerebral hypothermia during the latency and secondary energy failure periods include normalization (rather than decrease) of protein synthesis, reduction in free radicals and modulation of activation of microglia and cytokine production. Another key mechanistic finding for hypothermia’s potential as a therapeutic agent for hypoxic–ischemic brain injury was the reduction in the excitatory neurotransmitter glutamate when hypothermia was used compared with controls. In piglets, a 4°C temperature reduction attenuated the increase in excitatory amino acids and nitric oxide concentrations in brain extracellular fluid following hypoxia–ischemia .
With such promising results for a potential treatment initiated in the latency phase that could ameliorate multiple mechanisms of secondary energy failure that caused brain pathology in animals, investigators continued work to identify optimal dose and timing of cooling in anticipation of having to choose a degree and duration of cooling for clinical trials in human infants. The duration and degree of cooling have been shown to interact. In infant rats, a 2–3°C decrease in brain temperature for 72 h from the end of hypoxia prevented brain injury, while 6 h of cooling had an intermediate effect . In the same model, a greater reduction in body temperature, of 5°C for only 6 h, starting immediately after the insult, gave significant neuroprotection and neurobehavioral improvement 1 and 6 weeks later . However, in the current clinical setting, the immediate initiation of active cooling after perinatal hypoxic–ischemic injury is not feasible while assessing the degree of HIE, but strategies to avoid overwarming while assessing the encephalopathic newborn can be implemented without adding significant risk. For this reason, investigators explored how time of initiation of cooling after injury affected outcomes in animals, targeting the latent phase between primary and secondary energy failure. In the most cited series of experiments, Gunn et al. induced hypothermia in fetal sheep at various times after injury, and before and after the onset of predictable electrographic seizures in the animals, which occur between 5.5 and 6 h after injury . These studies revealed that moderate hypothermia induced by 90 min after reperfusion, and continued until 72 h after ischemia, prevented brain injury and improved electroencephalographic recovery . When the start of hypothermia was delayed until just before the expected onset of post-injury seizures in this animal model (5.5 h after reperfusion), partial neuroprotection was seen . With further delay until after seizures were established (8.5 h after reperfusion), there was no electrophysiological or overall histological protection with cooling . With the mounting evidence supporting a physiologic mechanism for hypothermia, and identification of optimal time of initiation, degree and duration of cooling in large animal studies, the initial infant trials of hypothermia for HIE were initiated.
The pilot trials of moderate hypothermia for HIE in term infants demonstrated acceptable safety using one of two modalities: selective head cooling  or whole-body hypothermia . The pilot studies led to the first two successful multicenter randomized trials of moderate hypothermia for HIE [32,60]. These two, plus two additional completed large multicenter trials (one recently presented and one published in the New England Journal of Medicine) are reviewed here  [ICE trial, Unpublished Data].
The industry-sponsored (Olympic Medical) multicenter randomized study of selective head cooling enrolled 243 infants with moderate or severe encephalopathy and an abnormal aEEG, who were either cared for using the combination of a cap with circulating water and an overhead radiant warmer to maintain a rectal temperature of 34–35°C for 72 h after study entry in the first 6 postnatal hours, or had temperatures managed by servo-control by an overhead radiant warmer with a goal to maintain the rectal temperature at 36.8–37.2°C. The primary outcome of the study was death or severe disability at 18 months. A total of 16 infants (7%) were lost to follow-up, but among those followed, death or severe disability occurred in 66% of infants randomized to conventional care and 55% randomized to the cooled group (odds ratio: 0.61; 95% CI: 0.34–1.09; p = 0.10). In accordance with the hypothesis of the lead investigators in this study, head cooling for infants with the most severe aEEG changes was not protective; however, head cooling to achieve a rectal temperature of 34–35°C for infants with less severe aEEG changes (n = 172) was protective, with an odds ratio of 0.42, and a 95% CI of 0.22–0.80 (p = 0.009) .
The NICHD Neonatal Research Network trial evaluated 102 infants randomized to hypothermia with whole-body cooling using a temperature-control blanket with circulating water to an esophageal temperature of 33.5°C for 72 h compared with 106 control infants randomized to conventional care. Enrolled infants met blood gas as well as neurologic exam criteria in the first 6 postnatal hours (TABLE 2). Skin temperature for the control group was maintained between 36.5 and 37.0°C with servo-controlled overhead warmers. Esophageal temperatures were monitored, but were not used to guide management in the control group. The primary outcome of the Network trial was death, or moderate or severe disability at 18 months of age (in the CoolCap trial, the primary outcome was death or severe disability). A total of 44% of infants in the hypothermia group died or survived with moderate to severe impairment at 18–22 month follow-up, compared with 62% of infants in the control group, with a risk ratio of 0.72 (95% CI: 0.54–0.95), for a number needed to treat to avoid the primary outcome of death or moderate to severe impairment of six. For the hypothermia group versus the control group, respectively, the disabling cerebral palsy occurred in 19.2 and 30.0%, for a risk ratio of 0.68 and a 95% CI of 0.38–1.22. Unlike the CoolCap study, where the benefit appeared to be limited to infants with moderate encephalopathy as defined by aEEG criteria, there was a trend for cooling to benefit infants in both moderate and severe encephalopathy groups noted in subgroup analyses. It is important to note that a prior aEEG was not included in the enrolment criteria for the Network Whole Body Hypothermia study. Eligibility was determined by biochemical and medical history parameters, and a baseline neurological examination. The eligibility and exclusion criteria are included in Box 1 . Two additional multicenter trials were recently completed. In the Total Body Hypothermia for Neonatal Encephalopathy (TOBY) trial, 325 infants were enrolled and randomized if they were at least 36 weeks gestational age, plus, at 10 min after birth, had either an Apgar score of 5 or less or a continued need for resuscitation, or, within 60 min after birth, metabolic acidosis (defined as any occurrence of umbilical cord, arterial or capillary pH of <7 or base deficit of ≥16 mmol/l). They also had to demonstrate encephalopathy and have at least 30 min of abnormal background activity or seizures on amplitude integrated electroencephalography. Cooled infants received 72 h of hypothermia to 33.5°C for 72 h with slow rewarming, much like the NICHD trial. The control group was maintained on servo-controlled radiant warmers with skin temperature probes, to maintain the rectal temperature at 37.0 ± 0.2°C. A total of 42 infants died and 32 survived with severe neurodevelopmental disability in the cooled group, and 44 infants died and 42 had severe disability in the noncooled group (p = 0.17). The rate of survival without a neurologic abnormality was significantly increased in the cooled group (71 out of 163 infants [44%] vs 45 out of 162 [28%] in the noncooled group; p = 0.003). Among survivors, cooling resulted in reduced risks of cerebral palsy (p = 0.03). The authors’ conclusions were that the induction of moderate hypothermia for 72 h in infants who had perinatal asphyxia did not significantly reduce the combined rate of death or severe disability, but resulted in improved neurologic outcomes in survivors. The authors noted the consistency of effect in improving survival free of disabilities among the three published trials .
For TOBY, three sets of criteria were required (A, B and C).
aEEG: Amplitude-integrated EEG; ICE: Infant Cooling Evaluation; NICU: Neonatal intensive care unit; TOBY: Total Body Hypothermia for Neonatal Encephalopathy.
The in-hospital outcomes and mortality results from a fourth pivotal trial, the Infant Cooling Evaluation (ICE) trial were reported at the Hot Topics in Neonatology annual conference in Washington DC (USA), in December 2008. In the ICE trial, infants were recruited from wide geographic areas, and were cooled on transport using HotCold gel packs cooled to 10°C. The cooled group core rectal temperature goal was 33–34°C for 72 h, and the control group rectal temperature goal was 36.7–37.3°C. Infants were excluded if they required supplemental oxygen with a fractional inspired oxygen (FiO2) greater than 0.80. Additional inclusion and exclusion criteria for the ICE trial are listed in Box 1. This study enrolled 221 infants between February 2001 and July 2007. Enrolment ended in July 2007 because investigators at study sites had lost equipoise following publication of the studies by Shankaran and Gluckman [32,60] and meta-analysis of these studies plus the pilot studies by Gunn et al.  and the study of whole-body hypothermia by Eicher et al.  demonstrating a consistent benefit . A 2-year follow-up of infants enrolled in the ICE trial was completed in July 2009. Pre-peer review results were promising, but we await further word on mortality plus morbidities from this fourth pivotal multicenter randomized study.
In all of the clinical trials, safety outcomes were tracked, including disorders of glucose homeostasis. The risks of hypothermia include hypothermia-induced insulin-resistant hyperglycemia, which has been noted in adults treated with hypothermia. Hyperglycemia alters brain tissue metabolism, increases neurotoxic glutamate and damages neuronal cell membranes . In addition to the threat of hyperglycemia, hypoglycemia may occur along with severe hypoxic–ischemic injury, as glucose is rapidly and anerobically metabolized [65,66]. The degree of hypoglycemia contributes to the severity of asphyxia injury in an animal model , and is associated with more severe encephalopathy  and accumulation of biomarkers consistent with cellular injury . Hypothermia may exacerbate hypoglycemia , but to date, no excess hypoglycemia or hyperglycemia attributable to use of moderate hypothermia in neonates with HIE has been reported from the clinical trials [60,62,70].
Azzopardi and Edwards conducted a speculative meta-analysis of the Eicher, CoolCap and NICHD studies [32,60,62–63]. They acknowledged the differences in the primary outcome among the three studies (Eicher reported 12-month outcome, and had high loss to follow-up; CoolCap reported death or severe disability; the NICHD study reported death or moderate and severe disability), and included them as ‘any’ disability. The cumulative results for 237 cooled infants and 241 controls were a relative risk (RR) of death or disability of 0.76, with a 95% CI of 0.65–0.89 .
Schulzke et al. reviewed five cooling trials involving 552 neonates. They identified the differences among studies – specifically, cooling techniques, the definition and severity of neurodevelopmental disability differences between studies, and the varying loss to follow-up. They included three studies for analysis of the primary outcome of mortality and disability at 18–22 months. These three studies included 223 cooled infants and 226 controls [57,60]. Hypothermia had a significant impact on the primary composite outcome of death or disability, with a RR of 0.78 and a 95% CI of 0.66–0.92. The estimated number needed to treat to have one infant avoid death or disability was eight .
Shah et al. reported a meta-analysis of four of the cooling trials, combining the results of Gunn, Gluckman, Shankaran and Eicher, with the primary outcome of death or impairment in childhood [32,58,60,62,63]. A total of 249 cooled infants and 248 control infants were included. There was a significant reduction in the risk of death or moderate to severe neurodevelopmental disability in infants who received hypothermia compared with control infants, with a RR of 0.76, and a 95% CI of 0.65–0.88. The number of encephalopathic infants to treat to avoid one infant to die or survive with moderate-to-severe disability was estimated to be six .
The fourth meta-analysis of hypothermia is included in the Cochrane registry, and was led by Susan Jacobs, who is also the principle investigator of the ICE trial. In the primary analysis, the Network and CoolCap trials, as well as the Eicher pilot study and the Gunn pilot study, were included, with 228 cooled and 251 control infants among the four studies. Their results were quite similar to the other reviews: typical RR of 0.76 with a 95% CI of 0.65–0.89, and a number needed to treat of seven .
All four meta-analyses included an assessment of whether or not treatment and safety effects were heterogeneic, and all found that the results of the included studies were homogenous – that is, findings of efficacy and safety of use of hypothermia in the clinical trials were similar and consistent. Assessments of secondary outcomes, including mortality separately, and disability separately, also demonstrated benefit. This is important because there were concerns that if mortality was averted by cooling, more survivors would be devastated. As there has been reduction in both mortality and disability alone, this does not seem to be the case. In the Network trial, mortality was reduced in the hypothermia group, 24% compared with 37% in the control group (RR: 0.68; 95% CI: 0.43–1.01), and the risk of disabling cerebral palsy was 19.2% for the hypothermia group and 30.0% among the control group for a RR of 0.68 and a 95% CI of 0.38–1.22 . In the more recently published TOBY trial, mortality was the same for cooled and regular care, but among survivors, disability was significantly decreased . Referring back to FIGURE 1, the results of the multicenter clinical trials indicate that the critical period for initiation of cooling appears to occur in the first 6 postnatal hours.
Prior to the studies and use of hypothermia for HIE, longitudinal evaluation of neurologic status over the first postnatal days to weeks was predictive of outcome [7,8]. Secondary analyses of data collected in the two pivotal hypothermia trials are likely to enable clinicians to have a better prognostic ability in the context of cooling. In the CoolCap study, information on severity of encephalopathy at randomization, and at day 4 was available for a subset of 177 infants. Infants in the cooled group and infants in the control group with grade 3 or the most severe HIE who improved their grade of encephalopathy by day 4 showed higher rates of favorable outcome that those that did not improve. All infants except three in the cooled group who had severe HIE at day 4 had an unfavorable outcome, and most infants (75%) with mild or normal status at day 4 had a favorable outcome in both groups. In both groups, infants with moderate HIE prerandomization who had an improved HIE grade on day 4 had a favorable prognosis, whereas infants who deteriorated did consistently poorly. Among infants whose moderate encephalopathy did not change over the 4 days, the cooled infants were more likely to end up with a positive outcome than controls (24 out of 31 cooled with good outcome vs 10 out of 29 controls; p = 0.002). This suggests that even if moderate encephalopathy persists after use of the CoolCap protocol, there is still a chance of improved outcome compared with prior epochs without hypothermia . This is consistent with the potential hypothermia effect of modifying secondary energy failure and the repair and remodeling that is induced by hypoxic–ischemic injury to the still-developing brain.
The TOBY and CoolCap trials, as mentioned previously, used aEEG criteria for enrolment, while the NICHD and ICE trials did not. aEEG is a bedside tool that can be used for cerebral function monitoring in term and near-term infants. The aEEG records a single channel EEG from biparietal electrodes, and the signal is then filtered, rectified, smoothed and amplitude integrated. The aEEG interpretation is based primarily on pattern recognition, and the aEEG correlates well with conventional EEG. Owing to its simplicity of operations relative to conventional EEG, it can be more feasibly used in the intensive care unit immediately after birth and throughout a cooling interval. One group has found 100% agreement between aEEG and conventional EEG findings in infants with HIE .
Spitzmiller et al. reviewed the evidence for EEG and aEEG predicting long-term outcome after hypoxic–ischemic injury and found several studies confirming the strong predictive value of conventional EEG. Patients with EEGs that improved in the first 12 postnatal hours were very likely to have normal outcomes . Several of the reviewed studies found that improvement in aEEG during the first postnatal days was also associated with good prognosis [77,78]. In the NICHD Network trial there was a trend toward benefit in neonates with both moderate and severe encephalopathy by physical examination criteria . There is no information currently available on whether whole-body hypothermia in the range and for the duration used in the NICHD Network trial has any potential benefit for infants demonstrating specifically moderate or severe encephalopathy aEEG criteria, but from the data collected in the NICHD study Shankaran et al. found that improvement in clinical examination severity of HIE and hypothermia treatment was associated with lower risk, and persistence of severe HIE was associated with a greater risk of death or disability following neonatal HIE . The CoolCap study had similar findings . The Network collected aEEG data on cooled infants, and we await results of their analyses assessing how cooling affects the aEEG trends, and how cooling influences the associations between early aEEG findings and neurologic outcome. Eventually, findings from combinations of alterations in MRI, aEEG, EEG and the evolving clinical examination may be useful in directed care or improving prognostic ability. The future could also include biomarkers that classify injury and help predict outcome, as they are showing promise in the rapid diagnosis of ischemic stroke in adults [80,81].
Both the Network and the CoolCap trials identified worse outcomes among infants with HIE who experienced high core body temperatures. In the CoolCap study, control patients had core temperatures monitored, and clinicians were directed to maintain rectal temperatures within the study range of 36.8–37.2°C. A total of 34 control patients had rectal temperatures of 38°C and over at any time during the 76-h monitoring period, 28 of whom had unfavorable outcomes; 76 control patients did not have pyrexia and 45 of these had unfavorable outcomes (OR: 3.2; 95% CI: 1.2–8.4; p = 0.028). Only 11 cooled patients (including one patient who was not cooled) had a rectal temperature of 38°C and over at randomization or at rewarming, and nine of these had unfavorable outcomes; while 50 of the 97 patients in the cooled group who did not experience pyrexia had unfavorable outcomes (OR: 4.2; 95% CI: 0.97–18; p = 0.11) . In the NICHD trial, esophageal temperature in the control group was monitored, but the goal temperatures for control infants in the study were those used for routine care, which has been servo-controlled skin temperatures maintained on radiant warmers, rather than direct measures of core temperatures. Of the 102 infants randomized to the usual care group, 50 infants had a maximum esophageal temperature over 38°C. Higher core temperatures were associated with significant increases in risk of death or impairment, with odds of death or disability increased 3.6- to fourfold for each 1°C increase in the highest quartile of skin or esophageal temperatures .
Cooling is a postnatal therapy for infants presenting with HIE. Initiation of cooling soon after a hypoxic–ischemic insult improves the chances of a good outcome in animal models of HIE [54,55]; initiation of cooling closer to the time of perinatal injury could improve outcomes. Although clinicians were optimistic that electronic fetal cardiotocography and fetal pulse oximetry might improve outcomes, neither one nor the combination has reliably provided obstetricians with definitive information that would lead to reducing the risk of hypoxic–ischemic injury via identification of infants that would benefit from operative delivery in the clinical trials of these monitoring techniques [83,84]. Although summary evidence in combined clinical trials has not suggested that these interventions (fetal cardiotocography and fetal saturation monitoring) will lead to changed outcomes, this may be in part because the prevalence of hypoxic–ischemic injury in the clinical trials of these monitoring techniques was low. The combined assessment of the standard fetal heart rate tracing with an automated analysis of the fetal electrocardiogram (ST-waveform analysis; STAN, Neoventa Medical, Moelndal, Sweden) has shown promise in reducing metabolic acidosis at delivery; however, this promising modality requires a combination of educational and technological resources for proper implementation before it can be accepted [85,86].
As the clinical trials of fetal oximetry and cardiotocography have been completed and results reported, and the use of novel techniques such as STAN monitoring and conventional fetal heart rate monitoring are discussed, obstetric practice changes. As obstetric practice changes, the incidence of moderate to severe HIE may change as well. In some regional epidemiologic reports, increasing ceserean section rates and less frequent post-term deliveries have been associated with a lower incidence of HIE-classified infants .
Even if changes in obstetric practice and improvements in monitoring techniques reduce metabolic acidosis, it is unlikely that all occurrences of hypoxic–ischemic injury will be prevented. We know from the cooling clinical trials that more than 40% of the cooled infants either died or survived with impairments; therefore, the search for effective interventions for neonates must continue beyond hypothermia. There are promising agents in the pipeline, but none at the stage of Phase III clinical trials.
Postnatal infusions of magnesium sulfate have been tested in a study, which enrolled 40 infants (20 treated with magnesium), with promising results. Magnesium may work by increasing extracellular magnesium concentrations to block neuronal influx of calcium enough to overcome the axonal depolarization associated with hypoxic–ischemia, but magnesium infusions may also produce unwanted hypotension [88,89]. Its use specifically for HIE in neonates requires further study.
Xenon, a noble gas that also acts as a N-methyl-d-aspartic acid antagonist, has been shown to have additive benefits to hypothermia when tested in a rodent model of brain injury, with benefit of the combination of hypothermia plus xenon resulting in greater short- and long-term neuroprotection than from either treatment alone . More recently, in a piglet model of HIE, xenon alone or xenon combined with hypothermia was associated with lower use of inotropes than in control animals or those treated with hypothermia alone . A pilot clinical trial of xenon and hypothermia is listed on ClinicalTrials.gov  but has not started enrolment as of this writing. Tested inhalation doses have been 50% inhaled xenon. If higher than 50% FiO2 is clinically required, lower concentrations of inhaled xenon will have to be used.
Recombinant human erythropoietin (rEpo) is neuroprotective in neonatal models of brain injury. Proposed mechanisms of neuroprotection include the activation of gene pathways that decrease oxidative injury, inflammation and apoptosis, while increasing vasculogenesis and neurogenesis . A Phase II study of rEpo for HIE testing doses in the range of those used to prevent anemia of prematurity has been completed by investigators in China. Death or moderate/severe disability occurred for 35 (43.8%) out of 80 infants in the control group and 18 (24.6%) out of 73 infants in the erythropoietin group (p = 0.017) at 18 months . The current US studies of rEpo for neuroprotection, a Phase II study for preterm infants  and a Phase I study for term infants with HIE , are currently underway.
At Duke University (NC, USA), we have started enrolment in a Phase I trial of autologous cord blood cells for HIE with infusions to begin before 14 postnatal days . Human cord blood cells contain multiple cell types, including hematopoetic and endothelial progenitor cells, and when administered in adult animals with hypoxia plus ischemic injury, they ameliorate injury and improve behavioral outcomes [94–96]. In one study of rat pups, human umbilical cord blood cells infused 24 h after hypoxic–ischemic injury migrated selectively to the injured area of the brain, and engraftment was associated with substantial alleviation of spastic paresis . Not every study in newborn rodents has found similar positive results  ; however, we believe there is a need to initiate Phase I trials of autologous cord blood for HIE in human infants with available autologous cells and high risk of poor outcome even with hypothermia, before non-evidence-based clinical use becomes prevalent.
The efficacy of hypothermia for HIE in term and near-term neonates with HIE has been consistently demonstrated in the multiple clinical trials to date. The effectiveness in clinical practice, and for longer term school-age outcomes, remains to be seen. Hypothermia should be used in concordance with the protocols used in the published randomized controlled trials, with initiation of cooling in the first 6 postnatal hours. Cooled infants must be closely followed for their neurodevelopmental outcomes. Results of the practice must be reported and institutions providing cooling must participate in registries or databases that will allow for future analyses of implementation and results of hypothermia in clinical practice . While early initiation of hypothermia (in the first 6 postnatal hours) becomes part of usual care, questions remain. Should we cool infants who are born in a referring hospital and appear to meet entry critera in the first hour, but on arrival at a tertiary center appear vigorous and have a normalized neurological exam? What should we do with infants who cannot be transferred to a tertiary cooling center by 6 postnatal hours? What should be done with infants who suffer a hypoxic–ischemic event in the first days after birth? These questions are being addressed, as in the Neonatal Research Network’s Late Hypothermia Clinical Trial , or will be addressed in ongoing and upcoming clinical trials. As we learn more about hypothermia, the evidence supports obstetricians and neonatologists partnering to advocate for the development of appropriate cooling protocols in regional centers that will develop expertise and maintain close neurodevelopmental follow-up of cooled infants and report results to large databases where effectiveness of hypothermia can be analyzed.
Given the consistency of results of the clinical trials, hypothermia protocols are increasingly implemented in tertiary care centers. The results of the ICE trial will influence whether or not standardized regimens for cooling on transport are initiated and whether or not relatively simple cooling techniques that could be operable in the developing world reduce mortality and morbidity from HIE. Further exploration of aEEG results will inform clinicians as to whether or not this tool adds to prognostic accuracy and should guide therapy before, during and after hypothermia. Additional trials of duration and degree of cooling, and whether or not cooling initiated after the first 6 postnatal hours will influence outcome, will be completed and further inform clinicians of the best use of this therapy. Trials of additional interventions targeting neonatal HIE are needed and must continue. We are optimistic that the safety and efficacy of additional treatments will be established. Given the experience of the hypothermia clinical trials requiring 2–3 years to complete enrolment and an additional 2 years to complete neurodevelopmental follow-up, it will be 5 years at the earliest that the new adjunct therapies will be ready for widespread clinical use. Alternative methodologies to assess neurologic outcomes earlier, or trials that include larger numbers of enroling centers, will reduce the wait. Hypothermia is likely also to be tested for its ability to provide neuroprotection for infants in the developing world, for more preterm neonates (33–35-weeks gestational age) with evidence of hypoxic–ischemic injury, neonates with heart disease requiring cardiac bypass surgery and neonates on extracorporeal membrane oxygenation.
Financial & competing interests disclosure
Seetha Shankaran has received support from NIH grant U10 HD21385. C Michael Cotten has received support from NIH grant U10 HD40492. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.
C Michael Cotten, Associate Professor of Pediatrics, Duke University Medical Center, Box 2739 DUMC, Durham, NC 27710, USA, Tel.: +1 919 681 4844, Fax: +1 919 681 6065, Email: ude.ekud.cm@010ettoc.
Seetha Shankaran, Professor of Pediatrics, Wayne State University School of Medicine, 3901 Beaubien Blvd, Detroit, MI 48201-2196, USA and Director, Neonatal-Perinatal Medicine, Children’s Hospital of Michigan and Hutzel Women’s Hospital, MI 48201-2196, USA, Tel.: +1 313 745 1436, Fax: +1 313 745 5867.
Papers of special note have been highlighted as:
• of interest
•• of considerable interest